The Record of Prodeinotherium in the Iberian Peninsula: New Data from the Vallès-Penedès Basin

Deinotheres (Proboscidea, Deinotheriidae) are a clade of non-elephantiform proboscideans that originated in Africa and dispersed into Eurasia by the early Miocene. In Europe, deinotheres are first recorded in Greece during MN3, although they did not become a common faunal element throughout Europe until MN4. Early Miocene (MN3–MN4) deinothere remains from Europe are generally assigned to a different species (Prodeinotherium cuvieri) than those from the early middle Miocene (Prodeinotherium bavaricum; MN5–MN6). In the Vallès-Penedès Basin (NE Iberian Peninsula), Prodeinotherium remains are very scarce and largely remain unpublished. To clarify their taxonomic assignment, we describe the available material and compare it with that from elsewhere in Europe. Based on size and a few diagnostic occlusal details, we tentatively recognize both Prodeinotherium cf. P. cuvieri and Prodeinotherium cf. P. bavaricum in the basin. Although all the studied sites had previously been correlated to MN4, the recognition of P. cf. P. bavaricum at els Casots and les Escletxes is consistent with ongoing litho- and magnetostratigraphic studies suggesting a slightly younger age for these sites. The lack of Prodeinotherium remains in older (MN3) localities from the Vallès-Penedès Basin, where Gomphotherium is already recorded, further supports the view that deinotheres dispersed into Western Europe somewhat later than gomphotheres.


Introduction
The Genus Prodeinotherium Deinotheres (family Deinotheriidae) are a clade of nonelephantifrom proboscideans characterized by a single pair of downwardly-recurved tusks and a lophodont cheek tooth morphology (Osborn 1936;Bergouniux and Crouzel 1962a;Harris 1973Harris , 1978Göhlich 1999;Sanders et al. 2010). Like proboscideans as a whole, deinotheres originated in Africa, where two subfamilies are distinguished (Sanders et al. 2004(Sanders et al. , 2010: the plesiomorphic and small Chilgatheriinae, with a single monotypic genus (Chilgatherium Sanders et al., 2004) from the late Oligocene of East Africa; and the larger and more derived Deinotheriinae, from the late Oligocene to early Pleistocene of Africa and the early to late Miocene of Eurasia, represented by several species.
The number of distinct deinotheriine genera is still a matter of debate. Some recent authors (Böhme et al. 2012;Pickford and Pourabrishami 2013) still distinguish a single genus (Deinotherium Kaup, 1829), but most other authors (e.g., Harris 1973Harris , 1978Gasparik 1993Gasparik , 2001Göhlich 1999;Huttunen 2002aHuttunen ,b, 2004Huttunen and Göhlich 2002;Duranthon et al. 2007;Markov 2008;Rasmussen and Gutierrez 2009;Sanders et al. 2010;Aiglstorfer et al. 2014;Konidaris et al. 2017;Alba et al. 2020) distinguish the smaller Prodeinotherium Éhik, 1930 (late Oligocene to early middle Miocene; but see Qiu et al. 2007 for a report of Prodeinotherium from the late Miocene of China) from the larger Deinotherium s.s. (late middle to late Miocene). The distinction between the two genera has been supported on dental, cranial, and postcranial features (Harris 1973(Harris , 1978Huttunen 2002aHuttunen , 2004Sanders et al. 2010). According to Harris (1973Harris ( , 1978, purported dental diagnostic features (other than size) between the two genera would include the more reduced postmetaloph ornamentation in M2 and M3, the presence of mesostyles in P3 and P4, and the longer and more recurved tusks in Deinotherium as compared with Prodeinotherium. Together with the more numerous cranial and postcranial differences noted by Harris (1973Harris ( , 1978, the aforementioned dental features have been accepted as diagnostic by many (e.g., Huttunen 2002a; Sanders et al. 2010) but not all subsequent researchers. In particular, Ginsburg and Chevrier (2001) noted that mesostyle development and postmetaloph ornamentation are too variable and inconsistent to distinguish the two genera, while purported differences in tusk shape and some cranial features would not hold when the European material is considered. An in-depth joint taxonomic revision of African, European, and Asian deinothere species would be required to re-evaluate the diagnostic features between the two genera, together with a rigorous cladistic analysis to further confirm that Deinotherium is monophyletic (instead of polyphyletic). Pending such studies, both taxonomic views seem reasonable, but given the undisputed differences in multiple postcranial features (Harris 1973(Harris , 1978Huttunen 2004), we tentatively follow most recent authors (e.g., Konidaris et al. 2017;Alba et al. 2020) in recognizing Prodeinotherium as a distinct genus.
Deinotheriids are first recorded in the late Oligocene of Africa by Chilgatherium at 28-27 Ma and subsequently by Prodeinotherium sp. at 27.5-24.0 Ma (Sanders et al. 2004(Sanders et al. , 2010Rasmussen and Gutierrez 2009), being only later recorded in Eurasia. The dispersal of proboscideans into Eurasia was once considered a single biochronological event in the framework of multiple intercontinental dispersal events that took place coinciding with the closure of the Tethys Seaway and the emergence of a terrestrial corridor (the so-called Gomphotherium Landbridge) in the current Middle East ca. 19 Ma, due to the collision of the Afro-Arabian and Eurasian plates (Rögl 1999;Harzhauser et al. 2007;Sen 2013). However, the formation of this landbridge was a multiepisodic phenomenon (e.g., Harzhauser et al. 2007), with several proboscidean lineages apparently having dispersed out of Africa at different times. Currently available data indicate that deinotheriids are first recorded in Asia than in Europe. In particular, Prodeinotherium is first recorded by the earliest Miocene (~ 22.5 Ma) in the Bugti area of Pakistan (Métais et al. 2009;Antoine et al. 2013), and only later recorded in eastern Europe (MN3, early Miocene; Koufos et al. 2003). Subsequently, this genus was present throughout most of the European Miocene from MN4 onward until it was replaced by Deinotherium (Göhlich 1999;Sanders et al. 2010;Konidaris et al. 2017).
Prodeinotherium bavaricum was originally erected as Dinotherium Bavaricum von Meyer, 1831, by von Meyer (1831: 297), although the species was not described until somewhat later (von Meyer 1833: pls. XXXIV figs. 12-15, XXXVI figs. 10-11, 16-17) based on material from Gmünd (currently Georgensgmünd, Germany, MN5). Von Meyer (1833) included the material from Chevilly in his newly erected species, whereas in contrast Depéret (1887) considered that the latter (whose authorship he dated to 1832) was a junior subjective synonym of D. cuvieri. Other authors (e.g., Lartet 1859;Mayet 1908Mayet , 1909 subsequently distinguished the two species but assigned to them specimens from overlapping time ranges-contrasting with the non-overlapping chronostratigraphic distribution currently recognized for them (see below). Although Lartet (1859) and Mayet (1908) mentioned distinguishing features between the two species, most of them have not withstood more recent scrutiny, even by authors who consider them distinct (Ginsburg and Chevrier 2001). This has led some authors to synonymize them, in which case P. bavaricum must take priority (Huttunen 2002a, b;contra Depéret 1887).
Two additional nominal species of Prodeinotherium from Europe were considered junior subjective synonyms of P. bavaricum by Huttunen (2002a, b): Prodeinotherium hungaricum Éhik, 1930, and Prodeinotherium petenyii Vörös, 1989. Ginsburg and Chevrier (2001 similarly considered P. hungaricum as a junior synonym of P. bavaricum, while Gasparik (1993Gasparik ( , 2001 considered P. petenyii as a junior synonym of P. hungaricum, which he considered as a valid species. Given their small size and the MN4 age of the type material of both P. petenyii and P. hungaricum, here we consider that both nominal species are junior subjective synonyms of P. cuvieri instead of P. bavaricum. Although P. bavaricum has sometimes been reported to coexist with Deinotherium during the middle Miocene (Bergounioux and Crouzel 1962b;Ginsburg and Chevrier 2001;Duranthon et al. 2007), this has been questioned or attributed to mixing of reworked faunas by more recent authors (Böhme et al. 2012;Pickford and Pourabrishami 2013;Konidaris et al. 2017). According to Konidaris et al. (2017), P. cuvieri is recorded during the early Miocene (MN3-MN4) and P. bavaricum during the middle Miocene (MN5-MN6).

Prodeinotherium in the Iberian Peninsula
Prodeinotherium remains are not abundantly represented in the Iberian Peninsula, at least according to the literature. All the remains reported by Bergounioux and Crouzel (1962b) in their review of Spanish deinotheres belong to Deinotherium, except for a few remains from the Vallès-Penedès Basin (see below). Azanza et al. (1993) reported a maxillary fragment with P4-M3 from Artesilla (MN4, Calatayud-Teruel Basin, Spain), which is correlated to local subzone Ca, with an estimated age of 16.5 Ma ( Van der Meulen et al. 2012). In turn, Antunes and Ginsburg (2003) reviewed the deinotheres from Portugal and reported 15 dentognathic specimens that they assigned to Deinotherium bavaricum. Most of the material derives from localities (Quinta das Pedreiras and Quinta da Barbacena) correlated to Tagus Basin Unit Va2 (MN4; Antunes 1996; Antunes and Ginsburg 2003; The NOW Community 2020). A tusk fragment from Quinta Grande (lower part of Unit Vb) and a mandible from Quintanelas (upper part of Unit Vb, based on Chelas 2) were correlated by Antunes and Ginsburg (2003) to MN5, but the former is more likely MN4 (The NOW Community 2020) whereas the latter might be MN4 or MN5. In light of the MN4 age for some of these sites, it is surprising that Antunes and Ginsburg (2003) assigned all the material to D. bavaricum, given that Ginsburg and Chevrier (2001) had distinguished the MN4 material from France as D. cuvieri. The latter was considered a synonym of D. bavaricum by Ginsburg in Antunes and Ginsburg (2003: 176)-which did not cite Ginsburg and Chevrier (2001). Given the arguments above, on biochronological grounds we consider that most (if not all) of the Portuguese material reported by Antunes and Ginsburg (2003) might belong to P. cuvieri-although this cannot be adequately ascertained based on the information provided by the authors.
In the Vallès-Penedès Basin, the taxonomic allocation of deinothere remains, which are scarce and mostly consist of fragmentary remains, is currently unclear. Solé Sabarís (1936) attributed an upper premolar from Can Corró (Caldes de Montbui) to Deinotherium giganteum Kaup, 1829, but the specimen was subsequently assigned to D. bavaricum by . The latter authors also reported D. cuvieri from Can Mas, and Crusafont et al. (1955) briefly described a lower molar fragment of D. cuvieri from les Escletxes (el Papiol). In turn, Bergounioux and Crouzel (1962b) assigned a partial m3 from Can Mas, seven teeth from les Escletxes del Papiol, and a dp4 from Can Canals to D. bavaricum (as they considered D. cuvieri was merely a synonym). Finally, the record of deinotheres at els Casots and el Canyet was first reported by Moyà Solà and Rius Font (1993) and Casanovas-Vilar et al. (2011b), respectively, but their species ascription remains to be determined. In summary, although part of the Prodeinotherium material from the Vallès-Penedès Basin has previously been reported, it has never been figured or described in detail.
Here we describe the available material and compare it with Prodeinotherium from elsewhere in Europe, to clarify its taxonomic identity.

Age and Geological Background
The Vallès-Penedès Basin (NE Iberian Peninsula) has yielded a rich fossil record of terrestrial vertebrates from the early to the late Miocene (Agustí et al. 1985;Casanovas-Vilar et al. 2016a, b). Early Miocene (Burdigalian, MN3-MN4) fossil sites are located in facies associated with small-range alluvial fan systems and lacustrine systems, the latter having delivered the richest fossil localities (Casanovas-Vilar et al. 2016a). Eustatic changes during the mid-Miocene Climatic Optimum led to three different marine transgressions (late Burdigalian, Langhian, and early Serravallian), and continental sedimentation was not fully re-established until later in the Serravallian (Casanovas-Vilar et al. 2016a).
The Prodeinotherium remains from the Vallès-Penedès Basin come from various sites ( Fig. 1) in the municipalities of el Papiol (Can Mas, les Escletxes, and prob-ably Can Canals), Castellbisbal (el Canyet), Subirats (els Casots), and Caldes de Montbui (Can Corró). Except for the last (whose age is unknown), all these sites have classically been correlated to MN4 (Agustí et al. 1985;Casanovas-Vilar et al. 2011a,b, 2016b. However, ongoing analysis of recently gathered litho-, bio-, and magnetostratigraphic data (Casanovas-Vilar et al. 2019; Jovells-Vaqué and Casanovas-Vilar in press; Jovells- Vaqué et al. 2017 preliminarily indicates that els Casots and especially les Escletxes are younger. Depending on the criteria used to define the MN4/MN5 boundary, these sites can be correlated to the uppermost MN4 or lowermost MN5. This boundary is commonly defined by the extinction of the eomyid Ligerimys Stehlin andSchaub, 1951 (e.g., Mein 1999;Kälin and Kempf 2009;Van der Meulen et al. 2012), which turns out to be a diachronous event in Western Europe, with the last Iberian species Ligerimys ellipticus Daams, 1976, surviving for about 0.5 Myr more than the last central European species, Ligerimys florancei Stehlin and Schaub, 1951 (see Van der Meulen et al. 2012). However, Van der Meulen et al. (2012) noted that the last occurrence of L. florancei in both areas is approximately simultaneous, so that Hilgen et al. (2012) redefined the boundary using this event and placed it at 16.4 Ma. This implies that many sites traditionally placed within MN4 (such as those belonging to Aragonian subzone Cb) are to be correlated to the earliest MN5.
Therefore, els Casots, which has only delivered scarce remains of L. ellipticus, is correlated to the MN5 with an estimated age of ca. 16 Ma (authors' unpublished data).
Ligerimys ellipticus is absent from les Escletxes, thus favoring a correlation with middle Aragonian zone D (Da?), early MN5 (see Van der Meulen et al. 2012). The younger age of this site is confirmed by its stratigraphical position, between a coquina layer and a carbonate reef platform, both assigned to the Langhian (early middle Miocene; Permanyer 1990). The age of Can Canals, Can Mas, and el Canyet is difficult to constrain as they have delivered very few or no small mammal remains at all. These sites are geographically very close to les Escletxes and belong to the same alluvial unit, which is certainly below the mentioned Langhian coquina layer and several meters above Molí Calopa, a classical MN3 site (Crusafont et al. 1955; see also Casanovas-Vilar et al. 2011b, 2016b. Therefore, their age may range between MN4 or early MN5. The scarce rodent remains recovered at el Canyet include the cricetid Democricetodon hispanicus (Agustí 1981; Jovells-Vaqué and Casanovas-Vilar in press), which rules out an age older than MN4.

Materials and Methods
The studied materials include 18 teeth housed at the ICP. The Can Corró specimen was not studied but included in the comparisons based on published data (Solé Sabarís 1936). Dental terminology generally follows Pickford and Pourabrishami (2013: figs. 2-3). Dental measurements of MD length and BL breadth were taken to the nearest 0.1 mm, and crown proportions were assessed by means of the BLI (computed as BL breadth/MD length × 100), bivariate plots of BL breadth vs. MD length, and box-andwhisker plots. The comparative sample includes figured specimens of Prodeinotherium from Europe (Table S1). For metrical analyses of the tooth loci recorded at the Vallès-Penedès Basin, given the potential confusion between Prodeinotherium and Deinotherium in MN6 (see above), we restricted the comparative sample to measurements of MN4 P. cuvieri and MN5 P. bavaricum from Europe taken from the literature (Éhik 1930;Vörös 1989;Azanza et al. 1993;Gasparik 1993;Ginsburg and Chevrier 2001;Sach and Heizmann 2001;Huttunen 2002b;Huttunen and Göhlich 2002;Antunes and Ginsburg 2003;Pickford and Pourabrishami 2013). Statistical comparisons between the two samples were done by means of t and Mann-Whitney tests using PAST (Hammer et al. 2001), and comparisons of Vallès-Penedès fossils with these samples were performed using z-scores.
Data availability All data generated during this study are included in this published article and its supplementary information files. The dataset of published measurements analyzed during the current study is available from the corresponding author on request.

Remarks
The genus Prodeinotherium was originally erected as Prodinotherium Éhik, 1930Éhik, , following Kaup's (1832 unjustified amendment of the original spelling of the type genus of the family (Deinotherium Kaup, 1829) as "Dinotherium," which was reinstated by Osborn (1936). On this basis, Harris (1973) emended the original spelling of "Prodinotherium" into Prodeinotherium. Such an emendation was indeed unjustified at the time-as it did not fit into any of the categories that would deem it incorrect according to the Code (ICZN 1999: Article 32). However, given that it has been in prevailing usage since Harris' (1973) emended diagnosis of the genus, currently it must be considered justified and is therefore adopted here, in agreement with other authors. The type species of the genus is Prodeinotherium hungaricum Éhik, 1930, generally considered a junior subjective synonym of one of the other species of the genus (e.g., Huttunen 2002a;Pickford and Pourabrishami 2013;Konidaris et al. 2017; but see Gasparik 1993Gasparik , 2001Markov 2008).

Description
The lower molars display a typical deinotheriid morphology, with a trilophodont m1 (Fig. 2M) and bilophodont m2 (Fig. 2O) and m3 (Fig. 2Q)-although only the distal half of the last tooth is preserved. The fragmentary m2 crown displays a long and marked prehypocristid that almost reaches the metalophid, and a very protruding distal cingulid. The m3 talonid is very large, distally protruding, and buccally tilted.

Remarks
The m1 morphology is consistent with both P. cuvieri and P. bavaricum (Table S1), whose proportions Table 1 List of Prodeinotherium cheek teeth from the Vallès-Penedès Basin with measurements (in mm). Parentheses indicate that measurements are estimated due to minor damage. Abbreviations: MD = mesiodistal length (mm); BL = buccolingual breadth (mm); BLI = breadth/length index (%), computed as BL / MD × 100. Actual measured values for IPS10480 are MD = 66.0 mm and BL = 63.5 mm. IPS12712 is labeled as coming from Can Mas, but most likely this is the same specimen as reported by Bergounioux and Crouzel (1962b) from Can Canals. IPS12708 was identified as an M1 by Bergounioux and Crouzel (1962b). Actual measurement for IPS12713 is BL = 40.5 mm. Data for Can Corró were taken from Solé Sabarís (1936) Table S4), which largely overlap and display no significant differences for this tooth (Tables S2-S3). However, the great distal projection of the distal cingulid in the el Canyet m2 most closely resembles some specimens of P. cuvieri, being more projecting than others and, particularly, of P. bavaricum (Table S1), where the cingulid is closer to the hypolophid. The large talonid of the m3 from el Canyet also more closely resembles specimens of P. cuvieri than P. bavaricum, although this feature is somewhat variable in both species (Table S1).
Prodeinotherium cf. P. bavaricum (von Meyer, 1831) Fig. 2B-E, G-L, N, R Material Seven cheek teeth (probably belonging to the same individual) from les Escletxes del Papiol, el Papiol; and four permanent cheek teeth and a deciduous premolar from els Casots, Subirats (see Table 1 for measurements and further details).
Description The DP3 from els Casots (Fig. 2B), which only lacks a small distal portion of the crown, is bilophodont and displays a vaguely figure-eight-shaped, very waisted contour (especially buccally) that is longer than broad, and broadest distally. The well-developed mesial cingulum is buccally flanked by a distinct parastyle, much larger than the mesiolingually located protostyle. The short postprotocrista shapes a slightly acute angle with the preprotocrista. The extensive central valley is lingually open, with a poorly defined lingual cingulum. Buccally, the ectoflexus is very marked, and the fine and low postparacrista and premetacrista do not constitute a conspicuous ectoloph; there is a small cuspule-like enamel thickening bucally to the end of the postparacrista. The short and steep postmetacrista is directed toward the damaged distal cingulum.
Based on their similar morphology and wear, the two P3 from els Casots (Fig. 2D-E) might belong to the same individual. They display a subsquare contour (slightly larger than broad), a slightly oblique mesial margin, and an indistinct ectoflexus. The protoloph is continuous (although the preprotocrista is not well defined close to the paracone), whereas the metaloph is discontinuous (because the posthypocrista does not surpass crown midline); the distally-oriented postparacrista is very short. The two lingual cusps are conical (not buccolingually compressed); the hypocone is more peripheral than the protocone. These cusps are separated by a deep transverse groove that opens into a V-shaped lingual notch, which is flanked by a moderately developed lingual cingulum. The mesial cingulum is well developed but there is no distinct parastyle, whereas the distal cingulum is poorly developed; the ectoloph ends in an indistinct metastyle that barely protrudes distally. The worn and partial P3 from les Escletxes (Fig. 2C), as far as it can be ascertained due to incomplete preservation, resembles the specimens from els Casots, except for the somewhat more convex distal profile.
The partial M2 from els Casots (Fig. 2G) displays a nontapering subsquare occlusal profile (minimally longer than broad), with two distally concave transverse lophs. The mesial and distal cingula are well developed, but the distal one is much narrower. The transverse valley is ample and buccally displays no entostyle. The postmetacrista is continuous with a faint convolute located on the buccal side of the crown. In turn, two partial upper molars from les Escletxes probably correspond to an M2 distobuccal crown fragment (Fig. 2H) and an M3 crown only lacking its mesialmost portion (Fig. 1I). The former specimen displays a less restricted median transverse valley than the latter, whereas the M3 displays a distally tapering contour (the metaloph is buccally more restricted than the protoloph). In both specimens, and unlike in the M2 from els Casots, the postmetacrista ends close to a marked (even if buccolingually narrow) convolute, which is close to the crown midline, just mesial to the marked distal cingulum.
The two p3s (probably antimeres) from les Escletxes (Fig. 2J-K) display a subtriangular (longer than broad) and mesially tapering contour, but their damaged condition precludes assessing several occlusal details. The ectoflexid is slightly obliquely oriented and appears continuous between protoconid and hypoconid. The metaconid (partially preserved in the left specimen) is well separated from the protoconid, but due to damage it cannot be ascertained if these cuspids were linked by a short transverse crest or whether a mesial tubercle (paraconid) was present. The median valley displays no secondary cuspids or cristids, and the entoconid is buccolingually compressed and located very peripherally. The p4 from les Escletxes (Fig. 2L) displays a subrectangular (longer than wide) contour, but the extent of the mesial valley and of the mesial cingulid cannot be assessed due to damage. The postprotocristid and prehypocristid constitute a continuous ectolophid that separates the median valley from the wide buccal notch, but the buccal cingulid cannot be ascertained. The distal cingulid is relatively well developed. The lower molars are only represented by a partial and digitally 'restored' specimen from les Escletxes (Fig. 2N), identified as an m1 due to its trilophodont occlusal pattern, as well as an m3 distal fragment from els Casots (Fig. 2R). The latter displays a buccally tilted and moderately developed talonid.

Remarks
The DP3 shows similar size and proportions to the few available specimens of P. cuvieri and, especially, P. bavaricum (Figs. 3F and 4F), and merely differs from the figured specimens of the latter species (Table S1) by the more pronounced interlophid waisting of the crown.
The described P3s fit well with the morphology of Prodeinotherium-peripheral hypocone, indistinct ectoflexus, weakly developed parastyle and metastyle, and mesial cingulum more marked than the distal-but show no particular similarities with either P. cuvieri or P. bavaricum (Table S1). The upper premolar from Can Corró, originally described as a P4 (Solé Sabarís 1936: fig. p. 84), is best identified as a P3 given its occlusal morphology and proportions, and resembles the els Casots specimens, except for the better developed lingual cingulum. Prodeinotherium bavaricum displays on average significantly longer P3s than P. cuvieri (Tables S2-S4), and the measurements for MD length of all the Vallès-Penedès specimens (Figs. 3A and 4A) rule out an assignment to the latter species (Table S4).
The M2? fragment from les Escletxes is not very informative but displays a similar distal crown morphology to the M3 from the same locality. The latter tooth was identified as an M1 by Bergounioux and Crouzel (1962b), but the development of the crown behind the distal lophid and the width of the metaloph (only subequal to the protoloph) is characteristic of an M3. The morphology of the els Casots M2 is consistent with that of both Prodeinotherium species (Table S1). Prodeinotherium bavaricum significantly differs from P. cuvieri by displaying on average absolutely longer and relatively narrower M2 (Tables S2-S3), and the dimensions of the els Casots specimen fit better with P. bavaricum (Figs. 3B and 4B), although the z-score does not reject an attribution to either species (Table S4). The morphology and BL breadth (Fig. 4C) of the M3 from les Escletxes is also compatible with both species (Tables S1-S4), which do not display significant differences in this regard.
As far as it can be ascertained due to damage, the two p3s from les Escletxes (Fig. 2J-K) fall within the range of occlusal variation of both P. cuvieri and P. bavaricum (Table S1). This tooth locus is quite variable, and unfortunately the only diagnostic feature (more reduced paraconid in P. cuvieri; Ginsburg and Chevrier 2001) cannot be ascertained in the studied specimens. Although P. cuvieri displays a significantly shorter p3 (Tables S2-S3), there is ample overlap and the p3 from les Escletxes fits well with both species (Figs. 3C and 4D; Table S4). Similarly, the p4 from les Escletxes resembles in morphology both P. cuvieri and P. bavaricum (Table S1). Although P. cuvieri displays significantly smaller p4 dimensions on average (Tables S2-S3), the MD length of the les Escletxes specimen fits well with both species (Fig. 4E; Table S4). Moreover, the morphology of the m1 from les Escletxes fits with both P. cuvieri and P. bavaricum (Table S1), even if it appears slightly more elongated than the m1 from el Canyet assigned to P. cuvieri; although the two species significantly differ in m1 BL breadth (Tables S2-S3), the size of the les Escletxes specimen agrees with both species (Fig. 4G; Table S4). Finally, in the els Casots m3, the talonid is less projecting and higher (almost cuspulid-like) than in el Canyet m3 (see above), more closely resembling the m3 of P. bavaricum (Table S1).
Prodeinotherium sp. Fig. 2A, F, P Material Two molars from Can Mas, and a deciduous premolar probably from Can Canals (or Can Mas, see remarks below), both sites at el Papiol (see Table 1 for measurements and further details).
Description The partial dp4 from Can Canals? ( Fig. 2A) preserves only two mesially-concave transverse lophids (the hypolophid and the somewhat narrower tritolophid). A loosely defined mesiodistal cristid originating from the hypoconulid approaches the hypolophid, partially blocking the distal transverse valley. The distal cingulid is moderately developed but buccolingually short.
A partial first molar from Can Mas (Fig. 2F) probably corresponds to an M1 buccal fragment, but it is too poorly preserved to assess any occlusal detail. The partial germ from Can Mas (Fig. 2P) is identified as an m3 because the hypolophid is somewhat narrower than the metalophid, but the development of the talonid cannot be ascertained.

Remarks
The partial dp4 described above, recorded in the ICP collections as from Can Mas, is in all probability the same specimen reported by Bergounioux and Crouzel (1962b) from Can Canals, which is likely the correct provenance. Both the dp4 and m3 BL breadth fit well with both species (Fig. 4I; Table S1).

Prodeinotherium Taxonomy and Evolution
Our comparisons of dental size and proportions (Figs. 3 and 4; Table S2) generally confirm Ginsburg and Chevier's (2001) assessment that P. cuvieri is somewhat smaller on average that P. bavaricum, but indicate that differences are only significant for some dental positions (and generally only for MD length; Table S3). Given these slight differences, it is not surprising that the two species were synonymized by Huttunen (2002a). However, our morphological and metrical comparisons further suggest that the two species differ in some unexpected ways-including not only a longer and relatively more elongated M2 in P. bavaricum on average, but also a shorter and relatively broader m3 in the latter species compared with P. cuvieri. Despite a considerable overlap, the latter difference appears particularly relevant because it goes against the general trend of dental size increase from P. cuvieri to P. bavaricum. This might be attributable to a concomitant reduction of the m3 talonidmore developed (i.e., distally-projecting) in Prodeinotherium than in Deinotherium (von Meyer 1833; Huttunen 2002a), but also less contracted (i.e., wider) in P. cuvieri than in P. bavaricum according to Lartet (1859). These differences in the m3 would lend further support to the distinct species status of P. cuvieri and P. bavaricum, supported by most current authors (e.g., Pickford and Pourabrishami 2013;Konidaris et al. 2017), and previously justified based on overall dental size differences and p3 occlusal morphology (Ginsburg and Chevrier 2001).
Providing an emended diagnosis for P. cuvieri and P. bavaricum is outside the scope of this paper because it would require a taxonomic revision of all Prodeinotherium species at an intercontinental level. The traditional recognition of three different species-P. bavaricum in Europe, Prodeinotherium pentapotamiae (Lydekker, 1876) in Asia, and Prodeinotherium hobleyi (Andrews, 1911) in Africawas largely based on geography (Harris 1978) and has not proven particularly useful for deciphering the evolution of the genus. It is generally recognized that P. hobleyi may be distinguished from European Prodeinotherium in several occlusal details (greater separation between the p3 protoconid and metaconid, relatively less elongate p4, and a more distinct tubercle on the m3 talonid ;Harris 1978;Sanders et al. 2010). However, P. hobleyi from the early Miocene appears smaller and somewhat more plesiomorphic (e.g., possession of more elaborated M2-M3 postmetaloph ornamentation, and presence of m1 ectostyle and p3 mesostyle) than younger material from the same continent (Harris 1978). In turn, both European and Asian Prodeinotherium species also appear somewhat more derived than their African counterparts in various occlusal details, whereas an evolutionary trend toward larger size apparently took place in parallel in Africa, Europe, and Asia from the early to the middle Miocene (Harris 1978). This pattern is overall consistent with Prodeinotherium originating in Africa and subsequently diversifying independently in several places of Eurasia. Unfortunately, it is not possible to determine to what extent morphological changes (other than size) occurred in parallel until the alpha-taxonomy of Prodeinotherium is further clarified. As in Europe, the differences in dental size and shape noted by Harris (1978) for P. hobleyi might justify the distinction of additional African species, and the same holds for Asia, where some authors (Tiwari et al. 2006;Qiu et al. 2007) distinguish several species-Prodeinotherium orlovii  (Sahni and Tripathi, 1957) and Prodeinotherium sinense Qiu et al., 2007-in addition to P. pentapotamiae. The comparisons reported here for the European material of Prodeinotherium will hopefully contribute to clarify the taxonomy of this genus when an in-depth revision is undertaken in the future.

Prodeinotherium from the Vallès-Penedès Basin
Unfortunately, the p3 occlusal morphology and the m3 proportions, which appear diagnostic between P. cuvieri and P. bavaricum (see above), cannot be adequately ascertained in the Vallès-Penedès sample, whose restricted and fragmentary nature recommends the use of open nomenclature in our species attributions. The great development of the m2 and m3 talonid in the el Canyet sample tentatively favors an assignment to P. cf. P. cuvieri. Conversely, the size of the els Casots M2, the more restricted talonid of the els Casots m3 and, especially, the MD length of the P3 from els Casots and Can Corró support an assignment to P. cf. P. bavaricum, while the samples from Can Mas and Can Canals? are best left unassigned to species. Given the range of these species elsewhere in Europe, our taxonomic identifications are consistent with a younger age for els Casots and les Escletxes, and an MN5 or MN6 age for Can Corró-thereby indicating that the early middle Miocene of the Vallès-Penedès Basin is better represented than previously assumed (e.g., Casanovas-Vilar et al. 2011b, 2016b. This indicates that the Vallès-Penedès Basin records the youngest Prodeinotherium remains from the Iberian Peninsula. The Vallès-Penedès record of Prodeinotherium is further consistent with evidence from elsewhere in the Iberian Peninsula (Antunes and Ginsburg 2003) and Western Europe (e.g., Göhlich 1999;Ginsburg and Chevrier 2001) indicating that the arrival of Prodeinotherium occurred later (MN4) than the dispersal of Gomphotherium (MN3; Tassy 1990; Bulot and Ginsburg 1993;Göhlich 1999;Antunes and Ginsburg 2003;Mazo and Van der Made 2012). Some authors (Mazo 1996;Mazo and Van der Made 2012) concluded that gomphotheres were not recorded in the Iberian Peninsula until MN4, but indeed Gomphotherium has been reported from the Vallès-Penedès Basin in Torrent de Vilaroc Crusafont et al. , 1955Crouzel 1957, 1958;Mazo and Van der Made 2012) and Sant Andreu de la Barca (Agustí and Galobart 1998;Casanovas-Vilar et al. 2011b, 2016b. The latter site is conclusively correlated to MN3 (Agustí et al. 1985), and in particular to Ramblian zone A (Casanovas-Vilar et al. 2011b, 2016b, while Torrent de Vilaroc must be slightly younger but similarly correlated to the same zone, given its stratigraphic position just a few meters above Molí Calopa (Crusafont et al. 1955). Based on the upper limit of zone A (Van der Meulen et al. 2012), an age older than 17 Ma can be inferred for both localities and, hence, for the dispersal of gomphotheres in the Vallès-Penedès Basin. In contrast, the oldest record of Prodeinotherium in the same basin could correspond either to the MN4 or earliest MN5. Els Casots and les Escletxes are unquestionably correlated to the early MN5 (as defined by Hilgen et al. 2012). Can Mas, Can Canals, and el Canyet are close to each other and located within the same alluvial unit, stratigraphically below les Escletxes and the marine Langhian deposits in this area. The scarce small mammal fauna recovered at el Canyet excludes an age older than MN4 for this site (Agustí 1981;Jovells-Vaqué and Casanovas-Vilar in press), and furthermore the three localities are several meters above the classical MN3 site of Molí Calopa. Therefore, they are certainly older than 16 Ma, but their correlation to MN4 or MN5 cannot be refined because of the lack of diagnostic small mammal data. However, the presence of P. cf. P. cuvieri at el Canyet is consistent with an older (MN4) age of this site as compared with les Escletxes, thus being chronologically similar to the other record of this species from Iberia, Artesilla (ca. 16.5 Ma; Azanza et al. 1993). Given the geographical location of the Iberian Peninsula, the proboscidean record from the early Miocene of the Vallès-Penedès Basin lends further support to the view that gomphotheres dispersed across Western Europe somewhat earlier than deinotheres.

Conclusions
The comparison of the Prodeinotherium remains from the Vallès-Penedès Basin with those from elsewhere in Europe indicates that two different species are recorded, which we tentatively assign to P. cf. P. cuvieri and P. cf. P. bavaricum. The presence of the latter taxon contrasts with the classical correlation of all the studied localities to the early Miocene (MN4) and supports recent studies that advocate for a somewhat younger age (early MN5) than previously assumed for most of them. In agreement with recent publications, our comparisons also support the distinct species status of P. cuvieri, and further indicate that, besides differences in overall tooth size, m3 size and proportions would be diagnostic between the two species. The record of Prodeinotherium in the Vallès-Penedès Basin further supports the view that deinotheres dispersed into western Europe somewhat later than gomphotheres.
2013-12470 to I.C.V), the Generalitat de Catalunya (CERCA Program and 2017 SGR 116), and the Operational Programme Research, Development and Education-Project "Postdoc@MUNI" (No. CZ.02.2.6 9/0.0/0.0/16_027/0008360 to À.H.L). We thank Jordi Galindo, Josep M. Robles, and Albert G. Sellés for technical assistance with the ICP collections, as well as two anonymous reviewers for constructive comments that helped to improve a previous version of this paper.
Funding The funding of this paper is reported in the Acknowledgments.

Data Availability
The fossil material studied in this paper is housed and adequately curated in the Institut Català de Paleontologia Miquel Crusafont, as is accessible by other researchers. All the data generated in the course of this study is provided in the paper or the supplementary material.

Conflicts of Interest
We declare no conflict of interest.