Isotopic evidence for omnivory among European cave bears: Late Pleistocene Ursus spelaeus from the Peştera cu Oase, Romania

Abstract

Previous bone collagen carbon and nitrogen isotopic studies of Late Pleistocene European cave bears (Ursus spelaeus) have shown that these bears frequently had low nitrogen isotope values, similar to those of herbivores and indicating either unusual physiology related to hibernation or a herbivorous diet. Isotopic analysis of animal bone from the Peştera cu Oase (Cave with Bones), Romania, shows that most of its cave bears had higher nitrogen isotope values than the associated herbivores and were, therefore, omnivorous. The Oase bears are securely identified as cave bears by both their morphology and DNA sequences. Although many cave bear populations may have behaved like herbivores, the Oase isotopic data demonstrate that cave bears were capable of altering their diets to become omnivores or even carnivores. These data therefore broaden the dietary profile of U. spelaeus and raise questions about the nature of the carnivore guild in Pleistocene Europe.

Late Pleistocene cave bear (Ursus spelaeus) diet and physiology is not well known, although based on skull, mandible, and tooth morphology, they have been inferred to have been largely herbivorous (1–4). Moreover, isotopic studies of central and western European cave bears have shown that, in most cases where their values are compared with herbivores and carnivores from the same site, they have δ¹⁵N values that plot with, or lower than, herbivores (Table 1). These dietary inferences are in the context of bears that had a period of dormancy, similar to modern black bears (Ursus americanus) and brown bears (Ursus arctos), given the abundance of hibernation nests in European caves. This pattern has been interpreted in several ways.

Table 1.

Previously determined isotope values of European cave bears, with the mean ± SD for N > 2

Site	Region	δ13C	δ13C range	δ15N	δ15N range	N	Age, cal ka*	Ref
Liñares	NW Spain	−21.1 ± 0.3	−21.7 to −20.6	3.0 ± 0.4	2.1–3.6	20	40	5
Eirós	NW Spain	−21.1 ± 0.3	−21.3 to −20.4	5.1 ± 0.3	4.4–5.5	8	29	6
Aldène	S France	−20.8 ± 0.4	−21.3 to −20.5	3.1 ± 1.5	2.1–4.8	3	<30	7
Mialet	S France	−20.2 ± 0.4	−20.7 to −19.6	2.8 ± 1.6	0.8–4.4	5	—	7
Olaskoa	Pyrenees	−20.5	−20.5	2.7	2.7	1	12	7
Chauvet	S France	−20.5 ± 0.3	−21.2 to −20.1	3.8 ± 0.9	3.1–4.8	16	27.3–42.1	8
Scladina-1A	Belgium	−22.1 ± 0.2	−22.5 to −21.8	4.9 ± 1.2	3.0–6.1	7	44	9
Scladina-4	Belgium	−23.2 ± 0.3	−23.6 to −23.0	3.6 ± 0.6	3.0–4.5	4	120–130	10
Divje Babe	Slovenia	−20.5 ± 0.4	−19.9 to −21.0	1.9 ± 0.7	0.6–3.3	11	52	11
Ramesch	Austria	−22.5	−22.5	2.4	2.4	1	45–48	12
Herdengel	Austria	−22.0, −22.1	−22.1 to −22.0	2.3, 2.5	2.3–2.5	2	41.4–43.8	12
Schwabenreith	Austria	−22.9	−22.9	2.5	2.5	1	55–121	12
Lieglloch	Austria	−20.1, −20.2	−20.2 to −20.1	2.8, 4.1	2.8–4.1	2	30–40	12
Hartelsgraben	Austria	−20.8, −21.0	−21.0 to −20.8	0.8, 3.4	0.8–3.4	2	32–47	12
Conturines	Italy	−22.3 ± 0.9	−23.3 to −21.6	2.0 ± 0.5	1.6–2.5	3	45.3–51.3	12
Potočka Zijalka	Slovenia	−20.4	−20.4	4.2	4.2	1	35	12
Vindija	Croatia	−20.7, −21.1	−21.1 to −20.7	1.3, 1.5	1.3–1.5	2	>34	13
Muierii†	Romania	20.4 ± 0.2	−20.7 to −20.2	5.1 ± 1.6	3.4–7.3	4	44–53
Cioclovina†	Romania	−20.3, −20.4	−20.3 to −20.4	3.1, 4.3	3.1–4.3	2	36, 45
Various	Europe	−21.5 ± 0.5	−22.0 to −20.6	8.4 ± 0.5	7.7–9.1	10	—	14‡

All isotope data are from bone samples derived from mature individuals; data from teeth and immature individuals are not included here or in Fig. 3. The individual values are provided in SI Table 3.

*Radiocarbon ages were converted to calendar years by using CalPal (www.calpal.de).

^†The Muierii and Cioclovina data are from OxA in the context of radiocarbon dating (15, 16).

^‡The individual various European specimens are from Hilderbrand et al. (14) and were provided by Hilderbrand (personal commmunication).

Based on comparisons of cave bear δ¹⁵N values to herbivore values from three Late Pleistocene sites in France and to modern bear isotope data, Bocherens et al. (7) concluded that the cave bears were herbivores. They ruled out the possibility that hibernation had an effect on δ¹⁵N values of modern bears, and by inference, it should not have affected cave bear δ¹⁵N values.

Fernández et al. (12) analyzed δ¹³C and δ¹⁵N data from central and western European cave bears and hypothesized that the δ¹⁵N values could be linked to climate. They found a link between δ¹⁵N values and temperature, with cave bears from colder environments having lower δ¹⁵N values. Although there are connections between climate and plant δ¹⁵N values (17) that show that colder environments often have lower δ¹⁵N values, they argued instead that the lower δ¹⁵N values observed in Alpine cave bears were linked to hibernation. They argued that, because dormancy may result in lower cave bear δ¹⁵N values, following Nelson et al. (11), colder environments require periods of longer dormancy that results in lower δ¹⁵N values.

An alternative possibility (18) is that they had a dietary preference for nitrogen-fixing plants, which have lower δ¹⁵N values. Such a diet might result in the very low δ¹⁵N values seen in some cave bear samples (Table 1). However, there are dietary constraints in terms of potential protein intake, feeding rates, and digestibility of plant matter for at least modern brown bears (19–21), suggesting that a dependence on such plants does not fully explain the cave bear δ¹⁵N pattern.

A fourth possibility, proposed by Hilderbrand et al. (14, 22) on the basis of higher δ¹⁵N values from a diverse sample of cave bears (Table 1), is that some cave bears were largely herbivorous, as are some modern bears, but that other cave bears were more carnivorous and hence exhibit δ¹⁵N values similar to those of many modern omnivorous U. arctos and U. americanus. In this interpretation, the variation in trophic level indicated by δ¹⁵N values would reflect ecological plasticity among U. spelaeus, in which they were more or less carnivorous depending on available food resources, body-size requirements, interspecific competition, seasonality, and temperature.

Regardless of the cause, 86.1% of the available individual European adult cave bear δ¹⁵N values (n = 101) are <6.0‰, and hence within the basically vegetarian range [the number rises to 95.6%, n = 91 without Hilderbrand et al.'s (14) 10 specimens] (Table 1). In the context of this distribution of δ¹⁵N values, routine isotopic assessment of a cave bear metapodial associated with an early modern human cranium in the Late Pleistocene Peştera cu Oase, Caraş-Severin, southwestern Romania (23–27) [supporting information (SI) Text and SI Figs. 4 and 5], provided an anomalously high δ¹⁵N value of 7.8‰ (SI Table 2, Oase O34-232). To further evaluate this result and address the cause of the generally low cave bear δ¹⁵N values, we measured the isotope values of cave bears and associated fauna from the Peştera cu Oase.

Affinities of the Peştera cu Oase Ursid Remains.

Because the primary concern of isotope analysis of Late Pleistocene U. spelaeus remains relates to the persistently low δ¹⁵N values obtained for them, in contrast to the variable but generally higher (more omnivorous) values generally found for U. arctos (7, 28–30), it is essential to establish the taxonomic affinities of the bear bones from the Peştera cu Oase.

Morphological Affinities.

A series of morphological contrasts between cave and brown bears are well established (31–34). These include, for U. spelaeus, broad molars with heavily wrinkled surfaces and additional cusps, loss of the first three premolars, pronounced doming of the forehead (especially in large males) (Fig. 1), large masseteric fossae and sagittal crests, large temporomandibular articulations, more robust long bones and metapodials, shorter extremities, pronounced tibial torsion, medial humeral curvature, and aspects of pedal facet morphology. All of these characteristics are found in all of the diagnostic bear remains from the Peştera cu Oase, and their morphology shows no distinctive differences from other cave bear samples that cannot be easily accommodated within models of normal, intraspecific temporal and spatial variation (25). There are no U. arctos in the sample.

Fig. 1.

Left anterolateral views of Peştera cu Oase adult U. spelaeus crania. From top to bottom, N32.1 (female), O34.62b (female), N32.21 (male), and N37.49 (male). The protuberance on the N32.21 dorsal neurocranium is a stalagmite.

Genetic Affinities.

To provide additional confirmation of the morphological evidence, mitochondrial DNA (mtDNA) was extracted, amplified, and sequenced from 19 ursid samples (SI Table 2). All 19 individual sequences of the Peştera cu Oase ursids show clear affinity to central European cave bear sequences (35) rather than to brown bears. They do not form a monophyletic group within cave bear mtDNA variation, and the range of the Oase bear haplotypes is spread throughout most of the variability known for central European cave bear populations from southern Germany, Austria, Croatia, and Slovakia (35–37).

Stable Isotopes of the Peştera cu Oase U. spelaeus Remains.

The Peştera cu Oase faunal isotope data are in Fig. 2 and SI Table 2. The herbivores (Capra ibex and Cervus elaphus) have values similar to those observed elsewhere for herbivores in Pleistocene Europe (38–40), with δ¹⁵N values ranging between 2.2‰ and 5.4‰. The higher Canis lupus δ¹⁵N values are similar to those for European carnivores (9, 10, 41, 42). Whereas the wolf and deer remains come from the same time period as the bears, the ibex remains are more recent. There is fluctuation in faunal δ¹⁵N values over the past 50,000 years (38), and it is not necessarily valid to compare fauna from different time periods. The variation observed in European fauna occurs principally at ≈13 thousand years before the present (ka BP) and in the Holocene, where herbivores have reduced δ¹⁵N values; this may have affected the ibex scapula but probably not the other samples.

Fig. 2.

Isotope data for cave bears and other fauna from the Peştera cu Oase, Romania. Filled diamonds, U. spelaeus; open squares, C. lupus; filled triangles, C. elaphus; open triangles, C. ibex. S-EVA values are used except for the specimens for which only OxA data are available.

The Oase cave bear isotope data (Fig. 2 and SI Table 2) are shifted negatively in δ¹³C values (-21.5 ± 0.4‰, n = 21), but in this they match the values of most of the other European cave bear samples (Table 1 and SI Table 3). However, the Oase bear δ¹⁵N values are principally between the relatively “carnivorous” values of 5.7–9.8‰, with one “herbivorous” value of 3.6‰ (7.8 ± 1.4‰, n = 21; 8.0 ± 1.1‰, n = 20 without the one low value). A Wilcoxon nonparametric comparison of the pooled herbivore versus cave bear δ¹⁵N values for the Oase sample provides a P < 0.001 (P = 0.001 if the more recent ibex scapula is deleted from the sample); P = 0.036 between the Oase bears and more carnivorous wolf δ¹⁵N values. These cave bear isotopic values cannot be the product of diagenesis, because the other fauna, including the contemporaneous canid and cervid remains, have expected δ¹⁵N values, and all have expected C/N values.

These Oase bear data contrast strongly with all previously published cases of cave bear δ¹⁵N values associated with herbivore data from the same site, in which the cave bears have δ¹⁵N values that are similar to, or lower than, herbivores from the same site. Most of the unpublished specimens from the Romanian sites of Cioclovina and Muierii (15–16) follow the same pattern (Table 1 and SI Table 3), although two specimens (Muierii 05-6 and 05-11) have δ¹⁵N values of 6.0‰ and 7.3‰, and three specimens from Scladina 1A [δ¹⁵N of 5.7‰, 6.0‰, and 6.1‰ (9)] are close (Fig. 3). The museum-derived cave bear sample analyzed by Hilderbrand et al. (14) produced δ¹⁵N values similar to the Oase ones, although that may be due in part to selection for large (“museum quality”) specimens (22) because the degree of carnivory (and hence δ¹⁵N values) tends to correlate with size, at least among North American brown bears (29).

Fig. 3.

Distributions of δ¹⁵N values for U. spelaeus remains from European OIS 4–2 Late Pleistocene sites (Table 1) and the Peştera cu Oase (SI Table 2). Samples: Liñares (1), Eirós (2), southern France and Pyrenees (3), Chauvet (4), Scladina 1A (5), Divje Babe (6), eastern Alps (7), Muierii and Cioclovina (8), European (9), and Oase (10).

There is a modest inverse relationship between δ¹³C and δ¹⁵N values in the Oase cave bears (r = −0.708, P < 0.001; without the one low outlier, r = −0.554, P = 0.011). This pattern is occasionally observed in isotopic studies, and it is often interpreted as a mixing line between two food sources (43). In this case, it would indicate that the cave bears had two distinct dietary protein sources, one being plant foods and the other being animals with low δ¹³C and high δ¹⁵N values. None of the Oase herbivore species could be the main source of animal protein for the cave bears, because the δ¹³C values are too different (there is almost no shift in δ¹³C value between prey and consumer). Whatever the source of the animal protein, it is apparent from the Oase cave bear isotope data that U. spelaeus, as with U. arctos, was not strictly vegetarian.

Discussion

The Peştera cu Oase isotopic results indicate that cave bears do not always have low δ¹⁵N values when compared with herbivores from the same site. Moreover, this pattern is evident, although less dominant, in samples from Muierii and Scladina 1A. Therefore, cave bears had the ability to consume animal protein and behaved like omnivores or even carnivores, even if in many cases they were herbivorous.

This inferred dietary flexibility for cave bears is evident in the stable isotope values within some samples, including Oase, Muierii, and Scladina 1A, as well as between samples. In this variation U. spelaeus closely parallels the range of isotopic values seen for modern European and North American brown bears (5, 7, 28–30) and the few data available for Late Pleistocene brown bears (10, 42).

As a consequence of these δ¹⁵N values, the dietary ecology of modern, higher-latitude bears (excluding polar bears) is relevant for that of cave bears, especially the North American brown bears (U. arctos, including the Kodiak and grizzly bears) given their high-latitude range, body-size variation, occupation of regions with less human ecological impact than most of Eurasia, and extensive database. Brown bear diets range from almost completely vegetarian, including ones with substantial amounts of fruit/berries, to ones containing a substantial amount of fish and/or ungulate meat (19–21, 29, 30, 44, 45). All aspects of their omnivorous diets have limitations in availability, potential feeding rates, and nutritional value in any given environment; adequate weight gain for survival, reproduction, and hibernation therefore depends on a mix of as many food resources as are available (19, 21). Meat consumption, in particular, varies widely among and within brown bear populations, due, among nonmaritime bears, to the availability of ungulate fauna (29, 30, 44, 45). Large adult males also appear to be more carnivorous than females or subadult bears (28, 29). North American black bears (U. americanus) appear to have similar plant/meat dietary proportions as brown bears (29), except that the larger brown bears are frequently more carnivorous when the prime meat is maritime (e.g., salmon) (46). This ecological flexibility of modern brown bears therefore makes an appropriate model to understand the range of isotopic values now available for European cave bears, both within and between site-specific samples.

The Oase bear isotope data also indicate that there is no need to invoke special species-wide effects, such as a dormancy effect or a reliance on specific plant foods, to explain the frequently low δ¹⁵N for European Late Pleistocene cave bears.

These isotopic data also imply that U. spelaeus participated, at least occasionally, in the European Late Pleistocene carnivore guild. This would have brought it into competition with the purer carnivores, especially cave lions and hyenas, as well as the more omnivorous ones such as brown bears and wolves. A very large bear obtaining much of its protein from vertebrate prey would have a serious effect on the dynamics of that guild. Moreover, contemporaneous European humans, both late archaic and early modern, consistently exhibit high δ¹⁵N values (13, 47, 48), and an additional large-bodied carnivore would have further intensified any human–carnivore interactions within the ecosystem.

Conclusions

Isotopic analysis of the Late Pleistocene cave bears from the Peştera cu Oase, Romania demonstrates that at least this bear population had omnivorous dietary adaptations. This is in contrast to most of the other published cave bear isotope data, which have been taken to show that they were herbivorous. The Oase data confirm that cave bears, as with modern brown bears, were omnivorous, although they may have frequently had dietary regimes that were primarily herbivorous. In an ecological context containing an abundance of large social carnivores, such as Late Pleistocene Europe, bears would have needed the dietary flexibility that is emerging from these stable isotope data. The U. spelaeus stable isotope data from the Peştera cu Oase, in conjunction with the data for cave bears elsewhere in Europe, now permit us to view them in a broader dietary framework and thereby raise questions as to their roles within the ecosystems of Pleistocene Europe.

Materials and Methods

The Peştera cu Oase Ursid Sample.

The overall assessment of the Peştera cu Oase ursid sample is based principally on those remains excavated from the Panta Strǎmoşilor, supplemented by surface materials in the other galleries of the Peştera cu Oase (see SI Text, SI Fig. 4 and 5). For isotopic and ancient DNA analysis, 22 ursid, and 10 other large mammals remains were selected from both the excavated and surface remains (see SI Table 2). Fifteen of the bear bones derive from excavations in the Panta Strǎmoşilor, whereas the others come from the surfaces of the adjacent galleries, some from hibernation nests. Other faunal species are rare within the Peştera cu Oase; sufficiently preserved bones for isotopic analysis include wolf (n = 4), ibex (n = 4), and red deer (n = 2). All of the specimens are bone and derive from fully mature individuals.

Given the mixed nature of the skeletal remains, there may be some duplication of individuals within species. The bear mtDNA haplotypes provide a minimum number of individuals of eight. The radiocarbon ages for the ibex provide a minimum of three individuals. The cervid hemimandibles, however, are right and left specimens, morphologically similar, may represent the same individual, and have similar isotopic values and radiocarbon ages; they are nonetheless plotted separately.

Morphological Assessment.

The faunal remains from the Peştera cu Oase were determined to skeletal element and species based on morphological characteristics. Apart from a few fragments, morphological and morphometric species identification was readily apparent for the >5,000 bear remains excavated in the Panta Strǎmoşilor and collected from the cave surfaces.

Sample Collection.

The analytical samples were derived from mature bone diaphyseal, rib, palatal, and basal mandibular cortical bone and removed by using a stainless-steel blade on a Dremel tool.

Ancient DNA.

Ancient DNA was extracted from 170 to 750 mg of bone material as described in Hofreiter et al. (36) and Rohland and Hofreiter (49, 50). Two hundred eight-five base pairs (bp) of the mitochondrial control region were amplified by using regular PCR with primers and amplification conditions from Hofreiter et al. (35) or under multiplex PCR conditions as described by Rompler et al. (51) by using primers from Hofreiter et al. (35, 36). Amplification products were cloned by using the TOPO-TA cloning kit (Invitrogen) by following supplier recommendations. Four to eight clones of each amplification product were sequenced on an ABI 3730 by using the BigDye Terminator v1.1 Cycle Sequencing kit and M13 universal primers. Clone sequences were aligned by using BioEdit (52) and merged to reconstruct the complete 285-bp sequence for each individual. Every sequence position was determined from clone sequences derived from at least two primary amplification products to avoid errors caused by ancient DNA template damage (53, 54). In cases of consistent differences among clones derived from independent amplification products, a third primary amplification was used to clarify the correct state for each position.

Stable Isotopes.

The key assumption behind the analysis of carbon and nitrogen stable isotopes is that bone collagen is formed throughout a mammal's life, and the carbon and nitrogen in collagen is ultimately derived from dietary protein (55–57). The isotope ratios of the two stable isotopes of carbon (¹³C/¹²C or δ¹³C) and nitrogen (¹⁵N/¹⁴N or δ¹⁵N) in dietary protein are commonly well preserved in European Late Pleistocene mammal bone collagen. Therefore, measurement of the δ¹³C and δ¹⁵N values of mammal bone collagen can be used to determine the main sources of dietary protein over the lifetime of that mammal. The bone collagen isotope values likely reflect dietary protein consumed over the past 10–20 years of life for an adult long-lived mammal (58). Although used to assess whether the dietary protein carbon is derived from marine versus terrestrial sources or from C₃ or C₄ plants, it is the trophic level of the protein consumed that is of interest here. Because the δ¹⁵N of bone collagen is ≈2‰ to ≈4‰ higher than dietary protein (59, 60), one can use it to determine the main protein sources in the diets of cave bears, especially compared with other fauna from the same site, to see whether dietary protein derived mainly from plants, as indicated by other isotopic studies of cave bears, or from higher trophic level protein sources. Of particular relevance here, experimental work with living bears (14) has shown that bone collagen δ¹⁵N values are an accurate reflection of diet.

A potential source of error for paleontological samples is postdepositional collagen bacterial and/or chemical diagenesis. Bone samples were therefore pretreated to remove exogenous carbon and isolate the original collagen. The resulting collagen extract is measured for a number of collagen quality indicators, including collagen yield, C/N ratio, percent C, and percent N in the collagen extract (61, 62).

We extracted collagen from 26 mammalian bones from the Oase site using standard extraction procedures [i.e., modified Longin (63) method (64) with the addition of an ultrafiltration step (65)]. Isotope measurements were made by using a ThermoFinnigan elemental analyzer coupled to a Delta Plus XP isotope ratio mass spectrometer at the Max Planck Institute for Evolutionary Anthropology, Leipzig, Germany. Carbon isotope ratio measurements (δ¹³C) were made relative to the vPDB standard, and the nitrogen isotope ratio measurements (δ¹⁵N) were made relative to the AIR standard. All samples were measured in duplicate, and the errors (1σ) are better than ±0.2‰.

Additional isotopic data produced at the Oxford Radiocarbon Accelerator Unit (ORAU) during sample pretreatment for radiocarbon dating were included for 12 specimens. The ORAU extraction procedures are similar to those that we used, including ultrafiltration [Brown et al. (65)]. All of the data reported here have C/N ratios that are within the accepted range of 2.9–3.6 (61).

In addition, 15 of the same specimens were AMS radiocarbon dated, principally by ORAU (OxA); two specimens were dated by the University of Vienna laboratory (VERA), one by using ultrafiltration and the other of an intact molar, which experience has shown provides results similar to bone dates run with ultrafiltration.