Evidence for parallel development of ever-growing molars in Early Pleistocene rodents from southern Spain and their paleoenvironmental implications

Agustí, J. and Piñero, P. 2023. Evidence for parallel development of ever-growing molars in Early Pleistocene rodents from southern Spain and their paleoenvironmental implications. Acta Palaeontologica Polonica 68 (2): 379–391. In this paper, we present a detailed survey on the rodent fauna from the site of Barranco de los Conejos (Guadix-Baza Basin, southern Spain). Its rodent fauna is composed of three arvicolines ( Orcemys giberti , Manchenomys oswaldoreigi , and Tibericola vandermeuleni ) and two murids ( Castillomys rivas and Apodemus atavus ). The three arvicoline species present ever-growing molars. Orcemys giberti and Manchenomys oswaldoreigi can be considered as descendants of local Mimomys species ( Mimomys medasensis and Mimomys tornensis , respectively), while Tibericola vandermeuleni is an eastern inmigrant. Loosening of roots in Orcemys giberti and Manchenomys oswaldoreigi is explained as an adaptation to a fossorial way of life, in relation to the Early Pleistocene glacial–interglacial dynamics, which led to cooler and drier conditions. This environmental change would also explain the dispersal of Tibericola from the eastern Mediterranean.


Introduction
The Early Pleistocene at 1.8 Ma is a key period in the evolution of the mammalian faunas in western Eurasia. At this time, strong glacial pulses forced by obliquity are recorded in the marine record for mid latitudes, coinciding with the onset of the Walker circulation at lower latitudes (Burckle 1995). This is also a time of significant dispersal events among mammalian faunas, such as the first record of Eastern ovibovine bovids of the genera Praeovibos and Soergelia . Almost coincident with this event took place the first hominin dispersal out of Africa to Eurasia, as recorded at Dmanisi site in Georgia at 1.8 Ma ago (Vekua et al. 2002;Lordkipanidze et al. 2007). Among the small mammals, the most significant event is the widespread dispersal in the Northern Hemisphere from east to west of the first arhizodont arvicoline rodents of the Allophaiomys (Allophaiomys deucalion, Allophaiomys plio caenicus ;Meulen 1973;Agustí 1991;Garapich and Nadachowski 1996). As a difference with the arvicoline rodents of the genus Mimomys, roots never appear, so the molars are ever growing. Development of ever-growing molars in this group has been interpreted as a response to a diet based on hard items, such as grasses. In this way, the dispersal of Allophaiomys has been interpreted as a result of the expansion of cold steppic conditions throughout Eurasia. However, this scenario has been challenged by the evidence recorded at the site of Barranco de los Conejos in the Guadix-Baza Basin. This site presents a peculiar arvicoline association, composed of three different arhizodont species, none of them referable to Allophaiomys: Orcemys giberti, Manchenomys oswaldoreigi, and Tibericola vandermeu leni. Orcemys giberti was originally described by Martin et al. (2018) from the sites of Barranco de los Conejos (type locality) and Barranco del Paso, both in the Guadix-Baza Basin. In this paper we have enlarged the lectotype of this species with new material, including m1, m3, and M2.
Manchenomys oswaldoreigi was described for the first time in Barranco de los Conejos. Tibericola vandermeuleni was originally described as a new species of Allophaiomys by Agustí (1991), based on nine m1 and one M3. In this paper, we have enlarged considerably the sample of this species, including four M3 and 14 complete m1, which provide a more complete overview of its variability. Finally, the murids Apodemus atavus and Castillomys rivas from Barranco de los Conejos are described for the first time in this paper.

Geological setting
The Barranco de los Conejos site is placed in the Guadix-Baza Basin, in southern Spain (Granada province; Agustí et al. 2013). This basin records a very complete continental succession ranging from the Late Miocene to the Middle Pleistocene (Hüsing et al. 2010;Agustí et al. 2015b;Piñero et al. 2018a). The paleontological record includes more than 60 fossiliferous levels, among them the sites of Barranco León and Fuente Nueva 3, which record the oldest evidence of hominin presence in Western Europe at 1.2-1.4 Ma (Toro-Moyano et al. 2013;Agustí et al. 2015a). As these sites, Barranco de los Conejos is included in the Upper Member of the Baza Formation (Oms et al. 2000;Agustí et al. 2013). Previous lithostratigraphic and magnetostratigraphic analyses have shown that Barranco de los Conejos is placed at the base of the upper Matuyama geomagnetic chron . From a biostratigraphic point of view, this site occupies an intermediate position between the earliest Pleistocene pre-Olduvai level of Galera 2 and the Early Pleistocene post-Olduvai site of Venta Micena (Oms et al. 2000Agustí et al. 2011). Provided that the age of Venta Micena has been estimated between 1.6-1.4 Ma (Duval et al. 2011;Agustí et al. 2015b), the age of Barranco de los Conejos can be estimated at about 1.8 Ma, roughly coeval with the Dmanisi site in Georgia.

Material and methods
The fossil material referred to here was collected from the Barranco de los Conejos site during several sampling campaigns, as it is also the case of the arvicoline material included in this paper (Mimomys sp., Mimomys medasensis). All the sediment retrieved during these campaigns was water-screened using superimposed 4-, 1-, and 0.5-mm mesh sieves. The rodent collection from Barranco de los Conejos includes 76 rodent teeth, corresponding to five different species. These fossils are housed at the Catalan Institute of Human Paleoecology and Social Evolution (IPHES-CERCA; Tarragona, Spain). All the measurements are expressed in millimetres and were taken with the software DinoCapture 2.0, using photographs from the Digital Microscope AM4115TL Dino-Lite Edge. Rodent teeth are illustrated by means of micrographs taken with Environmental Scanning Electron Microscopy (ESEM) at the Servei de Recursos Científics i Tècnics de la Universitat Rovira i Virgili (Tarragona, Spain).
The terminology and measuring methods employed in the descriptions of the arvicoline teeth (only m1 and M3 have been considered in the case of Manchenomys oswal doreigi and Tibericola vandermeuleni) are those of Meulen (1973), modified by Rabeder (1981) for Orcemys giberti (Fig. 1). Weerd (1976) was followed when we describe murid teeth, and length and width have been measured as defined by Martín-Suárez and Freudenthal (1993).  Orcemys giberti Martin, Tesakov, Agustí, and Johnston, 2018  Measurements.-See Table 1.
Description.-The m1 has a very simple pattern, composed of an anteroconid cap (AC2), five alternating triangles (T1, T2, T3, T4 and T5) and a posterior lobe. A thin deposit of cement can be recognized at the labial ends of the re-entrant angles. The AC2 is widely confluent with the T4 and T5, both triangles being also widely confluent. A prominent mimomyan ridge is always present, being the only mimomyan feature that can be recognized, with no evidence of enamel islet. The anterior wall of the AC2 is free-enameled as it is the case of the labial wall of the mimomyan ridge, characterized by high dentine tract. The connection between the T4 and the T3 is very straight, as it is the connection between the T3 and the T2. In contrast, the T2 and T1 are widely confluent. The connection between the T1 and the posterior lobe is again very straight.
The M2 has an occlusal pattern composed of an anterior lobe (AL1) and three alternating triangles (T2, T3, T4). The three triangles are disconnected, with no dentine connection between them.
The M3 shows a simple occlusal pattern, with an anterior lobe, two triangles (T2 and T3), and a rounded, elongated posterior cap. The anterior lobe is widely confluent with the T2, while the connection between the T2 and T1 is very straight. In contrast, the T1 is again widely confluent with the posterior cap. In the posterior cap, a very shallow BRA3 and a small LRA3 can be recognized.
Remarks.-The genus and species Orcemys giberti were established by Martin et al. (2018) on the basis of material coming from Barranco de los Conejos (type locality) and Barranco del Paso, both at the Baza Formation of the Guadix-Baza Basin. Previous papers assigned this large arvicoline to Mimomys sp. ). The original sample consisted in one m1 and one M3 from Barranco de los Conejos (holotype and paratype, respectively), five m1 (four of them broken or eroded), two broken m2, two broken m3, one broken M1/2, and two M3 from Barranco del Paso. Here we add new specimens from Barranco de los Conejos, which shed new light on the variability and phylogenetic relationships of this taxon. The morphology of the m1 occlusal surface led Martin et al. (2018) to raise the question as to whether this species could be included in the lagurine genus Borsodia. However, in the same paper this assignment was discarded since the Barranco de los Conejos teeth still retained a rest of cement in the re-entrant angles. Instead, Agustí et al. (2013) suggested that the new taxon could be an in situ descendant of Mimomys medasensis ( Fig. 2J-L), a hypothesis supported by the cladistic analysis presented in Martin et al. (2018). Described for the first time at the site of Islas Medas (NE Spain;Michaux 1971), Mimomys medasensis is a widely distributed species in the Iberian Plio-Pleistocene (Mein et al. 1978;Esteban Aenlle and López-Martínez 1987;Sevilla et al. 1991;Agustí et al. 2011;López-García et al. 2023). An increase in size and hypsodonty has been observed in this species through time (Chaline 1987;Sevilla et al. 1991). In contrast, the presence of this species in the Guadix-Baza Basin is scarce (Agustí et al. 2015b), restricted to the earliest Pleistocene of the Galera section (level Galera 1G; Agustí et al. 1997). In size, Mimomys medasensis is comparable to Orcemys giberti (see Table 1) although it differs because of the development of roots. However, some levels stratigraphically close to Barranco de los Conejos (Cortijo de Don Alfonso, Table 1. Measurements (in mm) of the teeth of Orcemys giberti from Barranco de los Conejos (this work); m1 of Mimomys sp. from Cortijo de Don Alfonso, Cementerio de Orce, and Fuentecica 5 (this paper); and m1 of Mimomys medasensis from Galera 1G (this work) and Almenara-Casablanca 1 (Aenlle and López Martínez 1987 Orce 2, Cementerio de Orce, Fuentecica 5; Agustí et al. 1987;Oms et al. 2000) present an advanced species of arvicoline whose characters are intermediate between the two species.
No formal description has been done of these specimens (see Fig. 2H, I), which in previous faunal lists were quoted either as Mimomys ostramosensis or Mimomys cf. ostramosensis (Agustí 1986;Agustí et al. 1987;Oms et al. 2000). This unnamed species is larger and more hypsodont than Mimomys medasensis, comparable in this way to Mimomys ostramo sensis. It is characterized by the persistence of mimomyan  Manchenomys oswaldoreigi (Agustí, Castillo, and Galobart, 1993) Tables 2, 3.

Measurements.-See
Description.-The occlusal pattern of the m1 is composed  of an anteroconid cap (AC2), five alternating triangles (T1, T2, T3, T4, and T5) and a posterior lobe. Cement is present in all the re-entrant angles. In all the three specimens the anteroconid is short and asymmetrical, the lingual wall being wider than the labial one. Enamel is always lacking in the anterior half of the wall of the anteroconid complex. The triangles are also asymmetrical, the lingual ones (T1, T3, T5) being wider than the labial ones (T2, T4). Specimens show undifferentiated or slightly negative enamel. LRA4 and BRA3 are not so deep, AC2 and the T4-T5 dentine fields being widely confluent. The T4 and T5 are alternating, being also widely confluent. Dentine channels between the posterior lobe, T1, T2, T3 and T4 are very narrow. The occlusal pattern of the only M3 is composed of a transverse anterior lobe, two alternating triangles (T2-T3) and a posterior cap (PC1). The anterior lobe, T2 and T3 are isolated, with no dentine channels. The T3 is widely confluent with the PC1. The PC1 is simple and rounded. A small T4 is recognized on the lingual wall of the PC1.
Remarks.-The species Manchenomys oswaldoreigi was initially assigned to the genus Mimomys, on the basis of the extended development of roots in the m3 (Agustí et al. 1993). Later it was included in the new genus Manchenomys, established on the basis of the species Manchenomys orcen sis from different levels at the late Early Pleistocene sites of Fuente Nueva 3 (type locality, Guadix-Baza Basin) and Quibas (Agustí et al. 2022b). The teeth from Barranco de los Conejos coincides both in size and shape with Manchenomys oswaldoreigi from its type locality, Gilena 2 (Tables 2  and 3), being smaller than the younger Manchenomys or censis. The values of the indexes A/L and B/W are also comparable to those of Manchenomys oswaldoreigi from Gilena 2, but also to those of Manchenomys orcensis from Fuente Nueva 3 and Quibas (Table 4). In contrast, the values of C/W in Manchenomys oswaldoreigi from Barranco de los Conejos are remarkably low, even when compared with the type species at Gilena 2, which may be an indication of a more archaic population.
Description.-In the m1, the occlusal pattern is composed of an anteroconid cap (AC2), five alternating triangles (T1-T5) and a posterior lobe. All the re-entrant angles are filled by abundant cement. There is a rounded, well-developed anteroconid cap. In four specimens the labial wall outlines a tiny BSA4. In a young individual, this BSA4 is fully developed, as it is a prominent LSA5 (Fig. 4). The LRA3 and BRA3 are very deep, so the dentine connection between the T4 and T5 is very straight, almost inexistent in some cases (see B/W values in Table 6). The connection between the AC2 and T5 is relatively wide (see C/W values in Table 6), but in two specimens it is also quite straight. All the specimens show undifferentiated or slightly negative enamel. In the M3, the occlusal pattern is composed of a transverse anterior lobe (AL1), three alternating triangles (T2, T3 and T4) and a posterior cap (PC1). The re-entrant angles are always filled by abundant cement. The T4 is widely confluent with the PC1. The T4 and T3 can be also confluent.
In contrast, the dentine connections between the T3 and T2, and between the T2 and AL1 are very straight. In all the cases, a well-developed LRA3 is present, which in some specimens can be very deep. ded in Allophaiomys by Agustí (1991). Later, the affi nities with Tibericola jordanica from Ubeidiya (Israel) became evident, particularly because of the very low C/W values, which enabled to differentiate this genus from the several archaic Allophaiomys species (Allophaiomys deu calion, Allophaiomys pliocaenicus, Allophaiomys ruffoi, Allophaiomys chalinei). Tibericola vandermeuleni from Barranco de los Conejos is certainly less derived than Tibericola jordanica, which presents a fully developed BSA3 and an incipient LRA5. This is in accordance for an older age of the site of Barranco de los Conejos with respect to Ubeidiya, a site dated to 1.4 Ma. A third Tibericola species, Tibericola sakaryaensis was described by Ünay et al. (2001) Heller, 1936  Material.-Two M1 (IPHES-BC-5, 11), and two m1 (IPHES-BC-6, 13). All from Lower Pleistocene, Barranco de los Conejos, Guadix-Baza Basin, Spain. Table 7.

Measurements.-See
Description.-In the M1, the t1 is slightly displaced backward with respect to the t2 and t3. The t1 has a posterior spur reaching basally the t4-t5 intersection in one specimen (IPHES-BC-5). The t2-t3 connection is higher than that of t1-t2. The t3 has a short posterior spur directed to the t5-t6 intersection. The t1bis and t2bis are absent. The well-deve lo ped, elongated t7 is connected to the t8, and separated from the t4. The t6 and t9 are connected. There is a reduced t12.
The m1 has a large, round tma. It is connected to the intersection of the two anteroconid lobes, forming a funnel in one specimen . The anteroconid complex is symmetrical. The lingual lobe of the anteroconid is connected to the protoconid-metaconid junction. There is no longitudinal crest. The labial cingulum is well developed. The oval posterior accessory cuspid (pac1) is similar in size to the tma. It is connected to the anterolabial face of the hypoconid. There are up to three additional accessory cuspids. The oval, lingually displaced posterior heel is variable in size.
Remarks.-The presence of t7 and t6-9 connection, the well-developed labial cingulum, and the absence of longitudinal crest in the studied specimens allow their ascription to the genus Apodemus. Apodemus gorafensis, Apodemus jeanteti, and Apodemus agustii are larger than the specimens from Barranco de los Conejos (Michaux 1967;Pasquier 1974;Ruiz Bustos et al. 1984;Martín-Suárez 1988;Bachelet 1990;Piñero et al. 2017;Agustí 2019, 2020;López-García et al. 2023). The studied material is close in size to Apodemus barbarae, but the presence of an individualized t7, and the complete union between the t6 and t9 in the M1 rule out ascription to this species (Weerd 1976). The size is also similar to that of Apodemus gudrunae, but again, the presence of a well-developed t7 precludes assigning the studied sample to this species (see Weerd 1976;Adrover et al. 1993;Piñero et al. 2018b). The small size, the presence of a large tma and the connection between the lingual lobe of the anteroconid and the protoconid-metaconid pair in the m1, and the separation between the t4 and t7 in the M1 are typical traits of Apodemus atavus (see Heller 1936). Furthermore, the material from Barranco de los Conejos is close in size to Apodemus atavus from Tollo de Chiclana According to several authors, Apodemus atavus and Apodemus dominans represent extreme phenotypes of a single species, Apodemus dominans being a junior synonym of Apodemus atavus (Fejfar and Storch 1990;Martín-Suárez and Mein 2004;Minwer-Barakat et al. 2005;García-Alix et al. 2008;Colombero et al. 2014). Some authors considered Apodemus atavus as the direct ancestor of the living Apodemus sylvaticus (Rietschel and Storch 1974;Fejfar and Storch 1990;Martín-Suárez and Mein 1998;Piñero et al. 2022). Genus Castillomys Michaux, 1969 Type species: Castillomys crusafonti Michaux, 1969, Layna, Pliocene. Castillomys rivas Martín-Suárez and Mein, 1991 Measurements.-See Table 8.
Description.-All the specimens have well-developed longitudinal crests, completing the connection among the tubercles of the crown. The M1 has the t1 displaced backward, and generally develops t1bis and t2bis. The posterior crests of the t1 and t3 are well developed and connected to the t4-t5 and t5-t6 intersections, respectively. There is no t7. The t4-t6, t9, and t8 are connected by a low crest. The t4-t8 crest is also low. The t12 is present as a small bulge between the t8 and t9.  In the M2, the t1 is connected to the t4-t5 intersection by a spur. The t1bis is present as a double t1. The round t3 is connected to the t5-t6 intersection by a narrow, low crest. The t7 is absent. The t4-t6, t9, and t8 are connected. There is a t4-t8 connection. The t12 is shown as a small salient between the t8 and t9.
The M3 has the t1 connected to the t5. The t3 is absent. There is a t4-t6 connection. The t9 is fused to the t8 forming a complex connected to the t6.
In the m1, the tma is absent. The anteroconid complex is slightly asymmetrical. The longitudinal crest is complete, being connected to the lingual part of the protoconid. The metaconid and entoconid are situated slightly anteriorly relative to the protoconid and hypoconid, respectively. The subtriangular or elongated posterior heel reaches the posterolingual base of the entoconid. The broad labial cingulum is separated from the protoconid by a valley. The oval or elongated pac1 is connected to the hypoconid by a spur. Another small accessory labial cuspid can be present.
In the m2, the large anterolabial cuspid is connected to the anterior side of the protoconid by a spur. The complete longitudinal crest may be connected to the metaconid-protoconid junction, or to the lingual side of the protoconid. The posterior heel may be oval or elongated. The labial cingulum is well developed and separated from the protoconid by a valley. There is a very small, low pac1 in three out of four specimens. No other accessory cuspids are present.
Remarks.-The small size, the well-developed longitudinal crest on the m1 and m2, the absence of t7, and the presence of a posterior crest on the t1 and t3 in the M1 and M2, are distinguishing features of Castillomys. The material from Barranco de los Conejos can be distinguished from Castillomys gracilis and Castillomys crusafonti by its larger size, and greater development of longitudinal connections both in upper and lower molars (Michaux 1969;Martín-Suárez and Mein 1991). The complete connection among the tubercles of the crown, and the presence of a broad labial cinculum separated from the protoconid by a valley in the m1 and m2 are features present in the species Castillomys rivas. In addition, the specimens from Barranco de los Conejos lie within the size range of Castillomys rivas from its type locality (Loma Quemada-1; Martín-Suárez and Mein 1991 The first appearance of the genus Castillomys presumably coincides with the beginning of the Pliocene (Weerd 1976;Mein et al. 1990;Piñero and Agustí 2019;Piñero et al. 2018a, whereas it disappeared at the Early-Middle Pleistocene boundary (Agustí et al. 1999). Martín-Suárez and Mein (1991) proposed the anagenetic evolutionary lineage Castillomys gracilis-Castillomys crusafonti-Castillo mys rivas, which underwent an increase in size and better development of the longitudinal connections along the Pliocene and Early Pleistocene.

Discussion
As a difference with other European sequences, in the Lower Pleistocene of southern Spain the first record of arvicolines with ever-growing molars is not represented by species of the genus Allophaiomys, but by three independent lineages, represented by Orcemys, Manchenomys, and Tibericola. While Orcemys and Manchenomys can be rooted in local populations of Mimomys (Mimomys meda sensis, Mimomys tornensis; both species present at the site of Almenara-Casablanca 1, in eastern Spain; Agustí et al. 2011; Esteban Aenlle and López-Martínez 1987) which developed ever-growing molars, Tibericola vandermeuleni is an eastern immigrant in the Guadix-Baza Basin. However, these three taxa are almost coeval with the first representatives of Allophaiomys (Allophaiomys deucalion), suggesting that the same environmental constraints that led to the origin of Allophaiomys were also responsible for the loss of roots in Orcemys, Manchenomys, and Tibericola. The causes for this parallel evolution have been a matter of discussion (Agustí et al. 2022b). The evolution towards development of evergrowing, hyperhypsodont molars in large herbivores has been classically explained on the assumption of a grassbased diet (Martin 1984;Janis 1988). However, among voles an alternative explanation has been proposed (Maul et al. 2014;Agustí et al. 2022b), since many representatives of this group display fossorial habits, as an evolutionary response to the glacial-interglacial dynamics. Voles use their incisors to burrow their galleries, so the teeth are exposed to heavy abrasion because of grit (Martin 1993), which at the end led to the development of ever-growing molars.
In any case, the evolution towards developing ever-growing molars is indicative of an environmental change that forced this trend in the Early Pleistocene voles of southern Spain. Evidence from other small vertebrates at this site such as amphibians and squamates revealed a change from the pre-Olduvai levels of the Galera section (Galera 2, Guadix-Baza Basin), characterized by warm and rather humid conditions, towards drier, less humid, seasonal conditions at Barranco de los Conejos ). Among insectivores, the absence of Crocidurinae at this site can be interpreted in the same way. This is also confirmed by the clear dominance of the murid Castillomys ri vas, a taxon linked to open grassy areas, against the scarce representation of Apodemus atavus, more related to forests (although Apodemus could have also consumed abrasive plants; Gomes Rodrigues et al. 2013). The entry at this time of eastern immigrants, such as Tibericola vander meuleni and the ovibovine bovids of the genus Praeovibos is also indicative of the spread to the west of the more arid conditions prevailing during the Early Pleistocene in the Mediterranean Levant. Therefore, the endemic rodent fauna from Barranco de los Conejos can be interpreted as a local response to the Early Pleistocene glacial-interglacial dynamics.

Conclusions
The site of Barranco de los Conejos presents a peculiar rodent association, characterized by three arhizodont arvicoline species: Orcemys giberti, Manchenomys oswaldoreigi, and Tibericola vandermeuleni. This association strongly differs from other coeval Early Pleistocene sites from Europe, such as Schernfeld, Villany-5, Betfia-2, Kamyk, and Mas Rambault (Kowalski 1960;Chaline 1972;Meulen 1974;Garapich and Nadachowski 1996), which are characterized by the presence of the first representatives of Allophaiomys (Allophaiomys deucalion, Allophaiomys pliocaenicus). The Barranco de los Conejos sample includes two endemic species, Orcemys gib erti and Manchenomys oswaldoreigi, which probably derived from local representatives of the genus Mimomys (Mimomys medasensis, Mimomys tornensis). The only exotic element is Tibericola vandermeuleni, a species whose closest relatives have been found in the Early Pleistocene sites from Turkey and Israel. The presence of Tibericola at Barranco de los Conejos is most probably the consequence of a dispersal event from the eastern to the western Mediterranean, an event which also included the first ovibovines of the genus Praeovibos. The trend towards loosening of roots in Orcemys and Manchenomys can be explained as an adaptation to the new environmental conditions forced by the glacial-interglacial dynamics of the Early Pleistocene, which in the case of the lower latitudes of southern Spain led to drier and cooler conditions, as it is also indicated by the prevalence of the murid Castillomys rivas against Apodemus atavus at Barranco de los Conejos. This change in the climatic conditions was most probably responsible for the western dispersal of eastern elements such as Tibericola and Praeovibos.