Effects of Global Warming on Ancient Mammalian Communities and Their Environments

Background Current global warming affects the composition and dynamics of mammalian communities and can increase extinction risk; however, long-term effects of warming on mammals are less understood. Dietary reconstructions inferred from stable isotopes of fossil herbivorous mammalian tooth enamel document environmental and climatic changes in ancient ecosystems, including C3/C4 transitions and relative seasonality. Methodology/Principal Findings Here, we use stable carbon and oxygen isotopes preserved in fossil teeth to document the magnitude of mammalian dietary shifts and ancient floral change during geologically documented glacial and interglacial periods during the Pliocene (∼1.9 million years ago) and Pleistocene (∼1.3 million years ago) in Florida. Stable isotope data demonstrate increased aridity, increased C4 grass consumption, inter-faunal dietary partitioning, increased isotopic niche breadth of mixed feeders, niche partitioning of phylogenetically similar taxa, and differences in relative seasonality with warming. Conclusion/Significance Our data show that global warming resulted in dramatic vegetation and dietary changes even at lower latitudes (∼28°N). Our results also question the use of models that predict the long term decline and extinction of species based on the assumption that niches are conserved over time. These findings have immediate relevance to clarifying possible biotic responses to current global warming in modern ecosystems.


Introduction
Recent global warming alters species distributions, abundances, interactions, and the timing of seasonal activities [1][2][3]. Bioclimatic 'envelope' models examining current warming trends predict the long term decline and extinction of species. Generally, these models are based on an understanding of the modern ecological parameters of species, and often incorporate an assumption of niche conservatism, i.e. the idea that ecological niches are maintained over long time scales [4][5][6]. The fossil record provides a long-term record from which the effects of past global warming can be assessed. Previous work comparing mammalian communities during the last ,780,000 years has documented only minor declines in small mammal species richness with increased warming [7]. Dietary reconstructions inferred from stable isotopes of mammalian tooth enamel yield minor to no changes between glacial and interglacial periods in both large and small mammals [8,9]. Similarly, there is evidence that the niches of mammalian taxa, based on temperature and precipitation, that persisted during the last glacial to interglacial transition, are conserved [10]. All of these studies suggest that mammalian responses to interglacial warming were generally minor.
Here we present stable carbon and oxygen isotope data of medium to large mammals from a glacial and an interglacial site in Florida. We first compare carbon isotope values and evaluate the hypothesis that dietary niches, inferred from the mean and breadth of carbon isotope values, did not change with interglacial warming. Next, carbon and oxygen isotope data from seriallysampled horse teeth are used to elucidate if and how relative seasonality differed between the glacial and interglacial period, with oxygen isotope data from all taxa documenting changes in relative humidity [11]. This study tests how interglacial warming affected mammalian diets as well as documents the magnitude of climatic differences at these low latitude (,28uN) glacial and interglacial sites.
Stable carbon and oxygen isotopes are incorporated into the tooth enamel of mammalian taxa and are representative, respectively, of the food and water consumed while alive. Furthermore, C 3 plants (e.g. trees, shrubs, and cool season grasses) photosynthesize differently from C 4 plants (e.g. warm season grasses) and subsequently reflect distinctly lower d 13 C ranges [12]. Taking into account the 13 C enrichment from food to tooth enamel (,14%) as well as the decline in d 13 C values (,1.5%) of atmospheric CO 2 due to fossil fuel burning over the past two centuries [13][14][15], tooth enamel values of less than 28% indicate a diet consisting of primarily C 3 vegetation whereas d 13 C values of greater than 22% indicate a diet of predominantly C 4 vegetation [12,14]. Additionally, C 3 grasses are a rare or absent component of the landscape in Florida during the last ,2 million years [16][17][18][19], further enabling d 13 C values of less than 28% to indicate a predominantly browsing diet. Lower d 13 C values can also indicate the consumption of browse in denser canopied C 3 forests [20][21][22]. By comparing stable carbon isotope values between individuals and populations during a glacial and an interglacial period, we can assess how long-term warming affected mammalian diets. Specifically, we compare the relative consumption of C 3 browse vs. C 4 grass, as inferred from d 13 C values. In addition, oxygen isotopes in mammalian tooth enamel typically document changes in temperature and precipitation, with 18 O enrichment indicating a warmer and/or drier climate [23][24][25]. Additionally, when comparing the oxygen isotope values of mammals that obtain a large proportion of their water from plants, relative aridity can be assessed [11,25].
Based on the depth of terrestrial fossils in relation to current sealevels, the coastal fossil sites of Inglis 1A and Leisey 1A ( Figure S1) represent a glacial and an interglacial site, respectively. Inglis 1A (29u09N, 82u419W; ,2.0-1.6 Ma) exhibits terrestrial fossils within a sinkhole in the Eocene Inglis Formation at depths of 5 meters below sea-level and completely lacks a marine fauna, indicating lower sea-levels that would have occurred during a glacial period [26]. This glacial interpretation is also supported by the presence of cooler adapted taxa such as the muskrat Ondatra cf. idahoensis and pronghorn Capromeryx arizonensis [26]. The younger Leisey 1A (27u429N, 82u309W; ,1.6-1.3 Ma) locality instead has terrestrial and marine fossils intermixed between two shell beds at depths consistent with interglacial levels [26]. Leisey 1A also has taxa indicative of a warmer interglacial period including alligators (Alligator mississippiensis [26]). These geographically similar terrestrial localities provide the rare opportunity to examine how mammals altered their diets in response to interglacial warming during the late Pliocene to early Pleistocene.
All mammalian taxa present within the orders of Artiodactyla, Perissodactyla, and Proboscidea were sampled for enamel stable carbon and oxygen isotopes (n = 115; Table S1 and S2; see supporting information and Materials and Methods). Representing 8 families, these taxa include the following: deer (Odocoileus virginianus), horses (Equus sp.), llamas (Hemiauchenia macrocephala, Palaeolama mirifica), peccaries (Mylohyus fossilis, Platygonus vetus), proboscideans (Cuvieronius tropicus, Mammut americanum, Mammuthus hayi), pronghorn (Capromeryx arizonensis), and tapirs (Tapirus sp., including Tapirus haysii). These mammals were compared both within (Table 1 and 2) and between fossil localities to demonstrate how dietary resources were partitioned. Serial samples (samples taken at a series of intervals perpendicular to the growth axis of the tooth) of the high-crowned horses were also taken from specimens at both localities, enabling comparisons of relative seasonality (n = 23; Table S3).

Dietary Partitioning
The glacial Inglis 1A locality represents a C 3 -dominated community with all taxa, except for Equus, having d 13 C values more negative than 29.1% (Figure 1 and S2, Table S1 and S2). As C 3 grasses are a rare or absent component of Florida landscapes during the last ,2 million years [16][17][18][19], all ungulate taxa are interpreted to be C 3 browsers, with the exception of Equus. Resource partitioning within this C 3 -dominated community is apparent with Odocoileus d 13 C values significantly less than all other taxa sampled (with n.1) and Tapirus significantly less than Platygonus and Hemiauchenia (Table 1). Equus is more enriched in 13 C than all other taxa (p,0.0001), indicating a diet consisting mainly of C 4 vegetation (i.e. warm season grasses, as C 4 dicots and CAM vegetation with similar d 13 C values are rare or absent in Florida [17,27,28]). Thus, this glacial site is dominated by C 3 browsers, although it is clear from the horse data that C 4 grasses were present. Due to the reliance of the majority of taxa on C 3 vegetation and the relative rarity of Equus at Inglis 1A, we hypothesize that C 4 grasses likely occurred in much lower abundance.
The interglacial fossil site Leisey 1A consists of taxa which display a large range in d 13 C values from 214.3 to 22.0% ( Figure 1 and S2, Table S1 and S2). All genera (with n.1) present at both Leisey 1A and Inglis 1A are significantly more enriched in 13 C at the interglacial site (Equus, p,0.05; Hemiauchenia, p = 0.0001; Odocoileus, p = 0.001; Platygonus, p = 0.0001) with the exception of Tapirus. Increased d 13 C values in these taxa demonstrates increased inclusion of C 4 grasses in their diets, although Odocoileus may instead be browsing from a more open canopied forest during the interglacial period. The lack of dietary changes in tapirs is consistent with their morphologically and isotopically inferred dense-canopy browsing diet over time [29,30]. Additionally, the mixed feeders, Hemiauchenia and Platygonus, increased their isotopic niche breadth from a total d 13 C range of 3.3% and 3.1% at the glacial site to 5.5% and 5.7% at the interglacial site, respectively ( Figure 1, Table S1). Leisey 1A potentially has increased ungulate diversity compared to Inglis 1A with the addition of the peccary (Mylohyus fossilis), the llama (Palaeolama mirifica), and two proboscideans (Cuvieronius tropicus, Mammuthus hayi), but lacks the pronghorn (Capromeryx arizonensis). However, Inglis 1A (,7700) has fewer specimens than Leisey 1A (,22,000) which may bias diversity estimates due to taphonomic processes. Both peccary taxa and both llama taxa show significantly different values. Mylohyus has lower d 13 C values than Platygonus (p,0.05; Figure 1, Table 2 and S1). Similarly, Palaeolama has low d 13 C values as compared to Hemiauchenia (p,0.0001; Figure 1, Table 2 and S1). Although the proboscideans cannot be statistically compared across sites due to limited sample size at Inglis 1A, the isotopically inferred C 4 grazing proboscideans (Cuvieronius, Mammuthus) are only present at Leisey 1A while the browsing mastodon (Mammut) is found at both localities. Because Mammuthus has one of its earliest occurrences in North America from Leisey 1A, it was absent from the older Inglis 1A locality regardless of ecological factors. However, Cuvieronius was present in the late Pliocene of North America [31] and could have been present at Inglis 1A; thus, its absence may represent an ecological signal. During the interglacial period represented by Leisey 1A, dietary resource use by the majority of ungulate taxa were significantly different from each other ( Table 2). However, the isotopically inferred browsers (Palaeolama, Tapirus, Mammut, and Odocoileus) lacked significant differences when compared to each other (Table 2). Similarly, the isotopically inferred grazers (Cuvieronius, Equus, and Mammuthus) lacked significant differences from one another ( Table 2). This high degree of dietary niche partitioning among the Leisey 1A mammalian community, especially among taxa within the same family, may contribute to its higher diversity ( Table 2 and S1). For example, Palaeolama and Mylohyus may be able to coexist with phylogenetically similar taxa during the interglacial period because they were able to successfully partition food resources. The increased inclusion of C 4 grasses by presumed dietary generalists such as Hemiauchenia and Platygonus, the presence of the C 4 grazers Cuvieronius and Mammuthus, and the relative abundance of Equus at Leisey 1A likely represents the increased abundance of C 4 vegetation and the potential expansion of C 4 grasslands during interglacial periods in Florida. However, despite the increased consumption of C 4 grasses by ungulates at Leisey 1A, none of the grazers are interpreted to consume only C 4 grasses based on the presence of individuals with d 13 C values of ,21.5%. The lack of obligate ungulate grazers during an interglacial period in Florida is somewhat surprising based on modern analogues and further demonstrates the importance of C 3 dietary resources for all taxa present during these time periods.
In contrast to our data, Koch et al. (1998) demonstrated minor to no differences in ungulates from a full glacial period to late glacial periods during the Pleistocene. The mastodon (Mammut americanum) population from the full glacial West Palm Beach locality (,25,000 BP) has significantly greater d 13 C values then at the late glacial Cutler Hammock locality (,11,000 to 9500 BP [8]). However, the remainder of the taxa lack significant dietary differences as inferred from d 13 C values [8]. The lack of significant changes in dietary resources may be due to the limited sample sizes of taxa from localities with radiocarbon dates that can be discretely defined as full glacial or late glacial ages. The scope of the Koch et al. (1998) paper was to compare taxa with more specialized diets of either browse or grass; therefore, by excluding potential dietary generalists such as Platygonus and/or Hemiauchenia, the effects of warming on dietary niches may be less apparent. Lastly, the effects of interglacial warming on mammalian communities and their environments may not have been as profound during the late Pleistocene to the early Holocene, as compared to the late Pliocene to the early Pleistocene.

Paleoclimate and Seasonality
Oxygen isotopes of the fossil mammals present at Leisey 1A have a greater range of d 18    collectively. Oxygen isotope values range from 23.4 to 0.3% at Inglis IA (total range of 3.7%) and from 25.1 to 2.9% at Leisey IA (total range of 8.0%; Figure 2 and S2, Table S1). For mammals that get the majority of their water from food, d 18 O values increase with temperature and/or aridity. Based on the aridity index of Levin et al. [11], the increased d 18 O range at Leisey 1A indicates a drier climate (Figure 2 and S2). These data are in agreement with 50,000-year-old pollen records that indicate drying with interglacial warming in Florida [32]. All genera (with n.1) present at both localities become significantly more enriched in 18 Figure 3, Table S3). The serial carbon isotope samples at Inglis 1A indicate greater variability than at Leisey 1A, with total d 13 C ranges of 2.3% and 1.3%, respectively. Aside from greater d 13 C variation at Inglis 1A, the d 13 C values oscillate in a predictable manner that correlates with seasonal warming and cooling (i.e. greater d 13 C values during the summer and lower d 13 C values during the winter; Figure 3). This oscillating pattern is likely the result of seasonal variability in the consumption of C 3 and C 4 vegetation and/or the 13 C enrichment and depletion of vegetation due to seasonal water stress [34,35]. Oxygen isotopes similarly track seasonal variation in temperature and/or precipitation at Inglis 1A. The greater d 18 [23,24]. Even once patterns of mineralization are accounted for [36][37][38], the pattern of d 18 O variation at the interglacial Leisey 1A site is similar to patterns of d 18 O variation in Florida today ( Figure  S3). These data further support the designations of Inglis 1A and Leisey 1A as a glacial and an interglacial site, respectively, demonstrating that changes in relative seasonality occur with increased warming even at low latitudes of ,28uN.

Concluding Remarks
Contrary to previous studies, we document dramatic dietary and floral changes with interglacial warming. The majority of taxa analyzed increase their mean d 13 C values and/or isotopic niche breadth with warming. Additionally, closely related taxa partitioned their dietary resources differently when sympatric at the interglacial site. Our data falsify the initial hypothesis of niche conservatism, instead showing that increased warming resulted in changes in both the type and breadth of resource use in mammals. Although d 13 C values reflect only an aspect of an animal's larger dietary niche, significant differences in d 13 C values as seen here, demonstrates considerable differences in a component of the dietary niches of mammalian taxa. These data imply that models which incorporate data under the assumption of niche conservation may not accurately predict the impacts of global warming on mammalian species. Furthermore, oxygen isotopes in fossil mammal teeth demonstrate increased aridity and decreased relative seasonality with interglacial warming. This study highlights the need for further investigations aimed at understanding paleoecology of species over various time and climatic scales for inferring the future effects of global warming.

Materials and Methods
A total of 115 specimens were sampled for stable isotopes of tooth enamel, the preferred tissue for geochemical analysis as it reliably reflects original isotopic values [12,14,39]. Late erupting teeth (e.g. fourth premolars and third molars) were preferentially selected for sampling when available; however, due to limited sample availability some early erupting teeth and/or fragmentary specimens were sampled. While the stable carbon and oxygen isotope values of early erupting teeth may be influenced by the consumption of the mother's milk, possibly resulting in differences in isotopic values, the early erupting and/or fragmentary teeth sampled here have isotopic values that are within the range of variation of late erupting teeth, and we therefore include these specimens in our analysis (Table S1 and S2). Using a low speed dental-style drill and carbide dental burrs, bulk samples were taken parallel to the growth axis of the tooth while serial samples were taken perpendicular to the growth axis. All enamel powder was pretreated with 30% hydrogen peroxide for 24 hours and 0.1 N acetic acid for 12 hours to remove organics and secondary carbonates, respectively [40]. Approximately 1 mg of these samples were then run on a VG Prism stable isotope ratio mass spectrometer with an in-line ISOCARB automatic sampler in the Department of Geological Sciences at the University of Florida. The analytical precision is 60.1%, based on replicate analyses of samples and standards . Stable isotope data were normalized to NBS-19 and are reported in conventional delta (d) notation for carbon (d 13 C) and oxygen (d 18 O), where d 13 C (parts per mil, %) = ((R sample /R standard )-1)*1000, and R = 13 C/ 12 C; and, d 18 O (parts per mil, %) = ((R sample /R standard )-1)*1000, and R = 18 O/ 16 O; and the standard is VPDB (Pee Dee Belemnite, Vienna Convention [41]). All stable isotopes are from the carbonate portion of tooth enamel hydroxylapatite.
All carbon and oxygen isotope values within the same locality were analyzed using Fisher's LSD multiple comparisons, as all samples from taxa with adequate sample size had d 13 C and d 18 O values that were normally distributed (Shapiro-Wilk tests). When comparing across genera between localities, t-tests were used. Ttests were also used to compare all individual serial samples per tooth, between localities. The aridity index of Levin et al. (2006) was used to test if interglacial warming resulted in increased aridity by comparing the total range of d 18 O values between localities. By comparing the d 18 O values of the entire fauna, the total range of d 18 O values of the most variable evaporation sensitive taxa (i.e. taxa that obtain a large portion of their water from plants) is captured.