Oxygen and Carbon Isotope Variations in a Modern Rodent Community – Implications for Palaeoenvironmental Reconstructions

Background The oxygen (δ18O) and carbon (δ13C) isotope compositions of bioapatite from skeletal remains of fossil mammals are well-established proxies for the reconstruction of palaeoenvironmental and palaeoclimatic conditions. Stable isotope studies of modern analogues are an important prerequisite for such reconstructions from fossil mammal remains. While numerous studies have investigated modern large- and medium-sized mammals, comparable studies are rare for small mammals. Due to their high abundance in terrestrial ecosystems, short life spans and small habitat size, small mammals are good recorders of local environments. Methodology/Findings The δ18O and δ13C values of teeth and bones of seven sympatric modern rodent species collected from owl pellets at a single locality were measured, and the inter-specific, intra-specific and intra-individual variations were evaluated. Minimum sample sizes to obtain reproducible population δ18O means within one standard deviation were determined. These parameters are comparable to existing data from large mammals. Additionally, the fractionation between coexisting carbonate (δ18OCO3) and phosphate (δ18OPO4) in rodent bioapatite was determined, and δ18O values were compared to existing calibration equations between the δ18O of rodent bioapatite and local surface water (δ18OLW). Specific calibration equations between δ18OPO4 and δ18OLW may be applicable on a taxonomic level higher than the species. However, a significant bias can occur when bone-based equations are applied to tooth-data and vice versa, which is due to differences in skeletal tissue formation times. δ13C values reflect the rodents’ diet and agree well with field observations of their nutritional behaviour. Conclusions/Significance Rodents have a high potential for the reconstruction of palaeoenvironmental conditions by means of bioapatite δ18O and δ13C analysis. No significant disadvantages compared to larger mammals were observed. However, for refined palaeoenvironmental reconstructions a better understanding of stable isotope signatures in modern analogous communities and potential biases due to seasonality effects, population dynamics and tissue formation rates is necessary.


Introduction
Stable isotope compositions of mammalian bioapatite are widely used proxies in palaeoenvironmental and palaeodietary studies. Starting with pioneering research on carbon [1] and oxygen isotopes [2][3][4][5] in the 1970s and 1980s, stable isotope analysis of bioapatite from fossil mammals rapidly became an established method for the reconstruction of palaeoclimate and palaeodiet.
For various reasons, especially to allow easier sampling and acquisition of sufficient sample material, most studies have focused predominantly on the isotopic analysis of skeletal remains from large mammals. New and improved mass spectrometric techniques allow oxygen and carbon isotope analysis of (sub)milligram sample amounts of bioapatite, bringing small mammal taxa such as small rodents (with a body mass below 1 kg) into the focus of interest.
So far only a few oxygen and/or carbon isotope studies are based exclusively or partly on fossil small rodents [8,[16][17][18][19][20][21][22][23][24][25][26][27]. However, for an improved understanding of oxygen and carbon isotope compositions in bioapatite of fossil small rodents, communities of modern analogues have to be studied extensively in order to investigate inter-and intra-specific variabilities. This has been done repeatedly for selected large mammals (e.g., [28][29][30][31][32][33][34][35]) but no comparable study has been conducted on small mammals so far. It is the aim of the present study to investigate such variations for small mammals. Such studies are fundamental in order to evaluate the number of individuals needed for statistically significant results and to ascertain whether speciesspecific calibrations or calibrations on a higher taxonomic level are suitable for the reconstruction of the oxygen isotope composition of local surface water. Furthermore, the intra-individual variability between bones and permanently growing teeth as well as the intrajaw variations need to be determined in order to investigate if these skeletal tissues record comparable isotope signatures or are seasonally biased.
We analysed the oxygen (d 18 O) and carbon isotope (d 13 C) composition (for definitions see Materials and Methods section) of bones and teeth from seven rodent species in multiple individuals. The samples derive from owl pellets of a locality in northwestern Germany, which were accumulated over a 4-year period (1991)(1992)(1993)(1994)(1995). Inter-specific, intra-specific, and intra-jaw variations were investigated, as well as variations between teeth and bones from the same individuals. These data were compared to published stable isotope data for large mammals.
Additionally, oxygen isotope analyses of coexisting carbonate (d 18 O CO3 ) and phosphate (d 18 O PO4 ) in rodent bioapatite were conducted on water voles (Arvicola terrestris). In order to determine if it is possible to make a correct estimation of the d 18 [4,16,36].
With the present study, the authors aim to provide a basic contribution for the enhancement of palaeoenvironmental interpretations of oxygen and carbon isotope data obtained from fossil rodents.

General Considerations
Oxygen isotopes. The oxygen isotope composition of mammalian bioapatite is determined by that of body water, which in turn is controlled by the oxygen isotope compositions of the different oxygen input sources (drinking water, food water, air oxygen, organic food compounds and water vapour in air) and oxygen output fluxes (exhaled CO 2 , liquid water in sweat, urine and feces as well as orally, nasally and transcutaneously released water vapour) [(e.g., [4,33,37]). The oxygen isotope composition of mammalian body water is linearly related to that of their drinking water (i.e., local surface water) for those mammals with an obligate drinking behaviour (e.g., [2,38]). Thus, empirical specific calibration equations relating d 18 O PO4/CO3 and d 18 O LW can be developed. Because the oxygen isotope composition of local surface water varies with air temperature and also with the amount of local precipitation and evapotranspiration (e.g., [39][40][41][42]), (palaeo-)climatic conditions can be inferred from the oxygen isotope composition of the skeletal remains of fossil mammals (e.g., [9] and references therein).
In bioapatite, oxygen is present in the phosphate, carbonate and the hydroxyl groups [6]. The average offset between d 18 O PO4 and d 18 O CO3 in skeletal apatite of different mammal taxa is around 9%. This offset ranges from 7.5 to 11.4% in the previously investigated modern taxa [43][44][45][46][47][48][49]. Both, d 18 O PO4 and d 18 O CO3 , can be used to track the oxygen isotope composition of ingested drinking water. For the offset between d 18 O PO4 and oxygen bound in the hydroxyl group, only a single calculated value of 216.6% is known [50].
Carbon isotopes. The carbon isotope composition of mammalian bioapatite is controlled by that of ingested food (i.e., plant material in herbivores) [1] and can therefore be used to investigate dietary preferences. Furthermore, the knowledge of the carbon isotope composition of ingested plant material can mirror specific habitats, type of vegetation cover (i.e., C 3 versus C 4 plants) and density, as well as ecological niches,feeding behaviour and seasonal changes of diet (e.g., [9] and references therein).
Significant differences in the carbon isotope composition of plants are induced by different carbon fixation strategies in plants, i.e., the C 3 , C 4 or CAM (Crassulacean Acid Metabolism) photosynthetic pathways (e.g. [51][52][53]). C 3 plants, which represent about 95% of all terrestrial plant species (e.g., [54]), have typical d 13 C values between -36 and 222% with an average value of 227%, whereas the d 13 C values of C 4 plants (mostly warm season grasses and sedges) have 215 to 210%, with an average value of 213% (e.g. [51,52,55,56]). CAM plants have highly variable d 13 C values within and between specific taxa, overlapping with both C 3 and C 4 plants (e.g., [55,57,58]). However, CAM plants represent only about 4% of all terrestrial plant species [54] and are mostly succulents [59], playing a minor role in the nutrition of most herbivorous mammals. Thus, a distinction between browsing and grazing herbivores based on bioapatite d 13 C values is possible, as shown by numerous palaeodietary studies (e.g. [60][61][62][63][64][65]). However, this approach is limited to ecosystems with coexisting C 3 and C 4 plants; the latter were not globally abundant until the late Miocene [66].
The carbon isotope composition of mammalian bioapatite is enriched by several per mil compared to the respective diet. Reported enrichment factor values range from 9 to 15%, depending on the digestive physiology of the investigated taxa (e.g., [80] and references therein).
Stable carbon and oxygen isotopes from bioapatite of fossil small rodents and case studies on modern material. Mammals evolved in the Late Triassic and retained relatively small body size until their radiation after the Cretaceous-Palaeogene transition [81]. Rodents first appear in the Palaeogene fossil record, and fossil rodent remains are widespread in Cenozoic terrestrial deposits, often in high numbers, accumulated predominantly by ancient avian predators. As reviewed by Grimes et al. [82], one of the main advantages of stable isotope studies on small mammals relative to large taxa is the higher abundance of small mammals in the fossil record, which enhances their availability. Furthermore, small mammals display a rapid evolution of their morphology, and hence small mammal remains are often index fossils for Cenozoic strata that enable a good biostratigraphic resolution. Additionally, most small mammals occupy a very restricted habitat, lacking long distance migratory behaviour, thus reflecting local palaeoenvironmental conditions more precisely than large mammals. Possible disadvantages compared to large mammals concerning the oxygen isotope composition may be a smaller proportion of drinking water in the overall oxygen intake, a more variable physiology (i.e., body temperature), and a possible stronger susceptibility to diagenetic alteration due to smaller, more fragile skeletal elements [82].
To date, very little is known about the inter-and intrapopulation variability between individuals from the same locality, as well as about variations between different teeth and bone material within a single individual. Only one study [83] has investigated intra-population and intra-jaw variations (d 18 O PO4 ) in small rodents based on a small number of samples of fat dormice (Glis glis, n = 3) from Great Missenden (Buckinghamshire, UK) and wood mice (Apodemus sylvaticus, n = 5) from Dungeness (Kent, UK). From these results, Lindars et al. [83] concluded that a quantity of.5 different post-weaning teeth should be used for isotope palaeo-thermometry. Further intra-population d 18 O PO4 data of multiple bioapatite analyses from small rodents (Apodemus flavicollis, Apodemus sylvaticus, Arvicola terrestris, Microtus arvalis and Pitymus sp.) are presented in a small quantity (n = 3-8) by D'Angela and Longinelli [38] and Longinelli et al. [36], but no data for distinct sympatric species from the same locality are available so far.

Material
The samples originate from fresh barn owl (Tyto alba) pellets accumulated over a maximum of four years, collected in April 1995 at ''Hof Gülker'', located in the nature reserve Rhader Wiesen, about 9 km north of Dorsten, North Rhine-Westphalia, Germany ( Fig. 1) and are part of the original material from Bülow [84]. Long term mean annual local temperatures in this region (IAEA-GNIP station Emmerich) average 10uC, and the mean annual precipitation is 744 mm. The long term weighted mean of the monthly d 18 O record of local precipitation between 1980 and 2005 is 27.3%. Considering only the period from April 1991 to April 1995 of owl pellet deposition, and hence the most likely time interval when the analysed rodents lived and mineralised their bones and teeth, the respective d 18 O value is 27.7. This is very close to the long term record [85]. The seasonal change in monthly precipitation of the sampling area is illustrated in Fig. 2.
The rodent skeletal material belongs to the four arvicolid species Arvicola terrestris, Myodes glareolus, Microtus agrestis and Microtus arvalis as well as to the three murid species Apodemus sylvaticus, Mus musculus, and Rattus norvegicus. From the latter species, the owl pellets contained only juvenile individuals. The analysed teeth were permanent growing (arvicolid and murid incisors and most arvicolid molars), thus reflecting the last four to twelve weeks in the life of the respective individual prior to death [86,87]. Bone material comprises stable isotope compositions over a longer time period, approaching the life span of the individual [e.g. 88], which in the present case is the time until predation by barn owls. None of the analysed taxa hibernate.  The isotope signatures of the skeletal tissue samples are likely to be seasonally biased, because of the different tissue formation and turnover periods and due to the seasonal population dynamics of the rodents and their predators. Population sizes of arvicolids and murids in temperate regions typically increase after a minimum at the end of the winter with the first breed in spring and reach a maximum (partly up to more than one order of magnitude larger than the early spring population) in late summer (e.g., [89][90][91]). Avian predator populations, responsible for pellet accumulations, are positively correlated to populations of their small mammal prey (e.g., [92][93][94][95]), causing a warm-season biased pellet deposition in modern ecosystems and probably in fossil ecosystems as well. Analogous conditions for modern and fossil samples are an important prerequisite for the development of d 18 O PO4/CO3d 18 O LW calibration equations and their application for palaeoclimate reconstructions [16].
No specific permits were required for the described field studies.

Methods
Oxygen and carbon isotope analyses of the carbonate in the bioapatite (d 18 O CO3 , d 13 C). Only teeth from taxonomically identified jaw specimens were used for stable isotope analysis. Teeth from each species, both upper incisors (in Arvicola terrestris and Rattus norvegicus only the left upper incisor) of ten individuals, were extracted from the jaw and inspected to ensure that no signs of digestive etching were present. Further d 18 O CO3 and d 13 C analyses were conducted on jaw bone material (derived from the zygomatic region) of five of the ten individuals of each species. In Arvicola terrestris, the left upper molars (M1-M3) of the ten specimens were also analysed. Bulk tooth and bone material was crushed and ground to a fine powder using an agate mortar and pestle. The chemical pretreatment procedure to remove organic matter and adherent carbonates followed Koch et al. [96]. Approximately 10 mg of sample powder were soaked with 30% H 2 O 2 (0.1 ml mg 21 ) for 24 h, rinsed five times with millipore water, soaked for another 24 h with an acetic acid -calcium acetate buffer solution (1 M, pH = 5, 0.05 ml mg 21 ) and rinsed again five times with millipore water, followed by drying overnight at 50uC. Typically, 1 to 1.3 mg of pretreated sample powder were reacted for 15 min. with 100% H 3 PO 4 at 70uC in a Thermo Scientific KIEL IV automated carbonate device. Released CO 2 was measured in dual inlet mode with a Finnigan Delta plus isotope ratio gas mass spectrometer at the stable isotope laboratory of the Geoscience Center at the University of Göttingen. The measured isotope compositions were normalised to the NBS 19 calcite standard, measured in the same runs together with the bioapatite samples. Data are reported in the d-notation in per mil (%), relative to the international isotope reference standards Vienna Standard Mean Ocean Water (VSMOW) for d 18 O and Vienna Pee Dee Belemnite (VPDB) for d 13 C [97].
where R sample and R standard are the 18 30.060.08% (1s) and a d 13 C value of -12.060.03% (1s) (n = 11). For inter-laboratory comparison, material of our internal AG-Lox enamel standard will be supplied upon request by the authors.
Oxygen analyses of the phosphate (d 18 O PO4 ). Analyses of the phosphate moiety of the bioapatite (d 18 O PO4 ) were conducted on the incisors, M1 and bone material from five of the Arvicola terrestris specimens. Ag 3 PO 4 was precipitated from about 4 mg of the pretreated sample powder (see section above), using a method slightly modified after Dettmann et al. [98] and described in detail by Tü tken et al. [8]. For dissolution of the samples, 0.8 ml of HF (2 M) were added, and the sample vials were put on a vibrating table for 12 h. This was followed by centrifugation and transfer to new vials of the supernatant sample solutions, leaving behind the CaF solid residue. After neutralising the HF solution with NH 4 OH (25%) in the presence of bromothymol blue as pHindicator, Ag 3 PO 4 was precipitated rapidly by adding 0.8 ml of 2 M silver nitrate (AgNO 3 ) solution. After settling of the small Ag 3 PO 4 crystals and centrifugation, the supernatant solution was pipetted off and the Ag 3 PO 4 was rinsed twice with 1.8 ml millipore water. The Ag 3 PO 4 was then dried overnight in an oven at 50uC. Ag 3 PO 4 aliquots of 0.5 mg were placed into silver capsules and analysed in triplicate by means of high temperature reduction Finnigan TC-EA coupled via a Conflo III to a Finnigan Delta Plus XL GC-IRMS, according to the method of Vennemann et al. [99].

Statistical Methods
The statistical analyses were performed using Wolfram Mathematica 7.0 and 8.0. To evaluate if the differences in the mean isotope values among multiple taxa are statistically significant, one-factor analyses of variance (ANOVA) were conducted, followed by posthoc-tests (Tukey) for pairwise comparisons. In one case, the assumptions of the parametric ANOVA were violated (non-normal distribution), therefore we additionally performed a Kruskal-Wallis non-parametric ANOVA that was not in disagreement with the parametric ANOVA results.
Minimum sample sizes to represent the population mean in d 18 O by 95% degree of confidence within 61s were determined by a standard bootstrapping approach according to the method presented by Fox-Dobbs et al. [100]. From the bulk incisor data sets of the seven analysed species (n = 10 each), 1,000 randomly selected subsamples (n sub ) were generated for every n sub between 2 and n-1. Then, the proportion of subsamples within all random replicates for a given n sub that range between 61s of the mean of the original datasets was determined, using the average value from 100 repeated evaluations. If the mean values of $950 out of 1000 subsamples are in a range between 61s of the analysed dataset, it is indicated that the respective n sub is adequate to estimate the population mean at least by a 95% confidence level.

Results
Oxygen and Carbon Isotope Data of the Carbonate in the Bioapatite (d 18 The data, which are summarised in Table S1,

Oxygen Isotope Composition of the Incisors
Intra-population variations in d 18 O CO3 from 2.0 to 3.9% have been observed in the bulk incisor samples (n = 10 of each species), with the exception of the arvicolid M. glareolus, which has a significantly higher range of 6.7%. However, if the most extreme value is excluded and considered as an outlier, M. glareolus has an intra-population variation of 3.5% that falls within the range of variation of the other species (

Carbon Isotope Composition of the Incisors
The  Table 1). Mean d 13 C values differed significantly between the analysed species (one-way ANOVA, F = 20.632, P,0.01). By pair-wise comparison, mean d 13 C values of two murids (M. musculus and R. norvegicus) differed significantly from those of all arvicolids as well as from A. sylvaticus. The mean d 13 C value of M. agrestis showed significant differences relative to those of A. sylvaticus and M. glareolus (Tukey test, P,0.05). Considering the arvicolids separately, differences in the mean d 13 C values between taxa (one-way ANOVA, F = 28.813, P,0.01) were observed. By pair-wise comparisons, significant differences between M. agrestis and M. glareolus themselves and with both other Arvicolidae were detected (Tukey test, P,0.05). Within murids, significant differences among taxa were present as well (one-way ANOVA, F = 7.230, P,0.01), but only A. sylvaticus and R. norvegicus show significant differences by pair-wise comparison (Tukey test, P,0.05).    (Fig. 5, Table 2).

Oxygen and Carbon Isotope Compositions of the Molars of A. terrestris (Intra-jaw Variations)
In A. terrestris the intra-jaw range (n = 10) between the molars (M1-M3) of the same individual is 0.2 to 1.0% in d 18 O CO3 and 0.1 to 0.6% in d 13 C. The overall intra-jaw range (I1, M1-M3) varies from 0.3 to 1.9% in d 18 O CO3 , while in most cases the incisors have higher values than the molars (Fig. 6, Table S1). In d 13 C, nearly all incisors have lower values than the corresponding molars, with overall intra-jaw variations between 0.3 to 1.5% (Fig. 7 (Fig. 8, Table 3).

Minimum Sample Size Calculations
Minimum sample sizes, calculated with the bootstrapping approach used by Fox-Dobbs et al. [100] are in the same range or smaller than previously published data from large mammals (e.g., [28][29][30]32]; Table 4). From all randomly chosen subsamples with n sub $4, 95% or more range within 1s of the mean of the complete data. In the case of a subsample size of n sub $7, even a 99% confidence is reached for a range within 1s of the mean of the whole dataset.

Inter-and Intra-specific Variations in d 18 O CO3
Mean d 18 O CO3 values of the analysed species (Fig. 3) show no, or only minor, differences. This suggests that probably all taxa drank from isotopically similar water sources. The higher d 18 O CO3 in R. norvegicus compared to the other rodent species may be attributed to the juvenile status of the individuals and might reflect a remnant suckling signal.
Due to the lack of published small mammal d 18 O CO3 data, the intra-population variability of the analysed species can only be compared to published d 18 O PO4 values of rodents. The ranges   [101], feral and domestic horses (Equus caballus) with 2.0 to 6.5% [28,30,35], and    bobcats (Lynx rufus) with 3.1% intra-population variability at a single locality [32]. Other published intra-population variabilities including three or more individuals for large terrestrial mammals are significantly larger than those observed in the present study, as is the case for mule deer (Odocoileus hemionus) with a range of 6.2%, coyotes (Canis latrans) with 6.3% [32], domestic yaks (Bos grunniens) with 3.7 to 8.5% and domestic goats (Capra aegagrus hircus) with 5.4 to 10.6% intra-population variability [30].

Inter-and Intra-specific Variations in d 13 C
Mean d 13 C values of the analysed species are in part significantly distinct from each other, notably between arvicolids and murids, indicating dietary differences (Fig. 3). The intrapopulation variations in d 13 C have a wide range from 1.8 to 12.7%, reflecting variations in the d 13 C values of their diet and in the nutritional behaviour (i.e., dietary specialists vs. dietary opportunists). Previously published data on intra-population variations from large mammals range from 0.7 to 8.5% [29,30,32,101].
The average carbon isotope fractionation between diet and bioapatite in rodents (determined on bone material) is+9.9% [1,23,102]. This suggests average diet d 13 C values between 229.6 and 226.1% for the arvicolids. For the murids, the average diet d 13 C values are higher: 225.8% in A. sylvaticus, 222.0% in M. musculus and 219.5% in R. norvegicus.
The low intra-population variations of A. terrestris, M. glareolus and M. agrestis (,2.2%) indicate little dietary variations, which is supported for A. terrestris and M. agrestis by the observation of nearly exclusively herbivorous diets [103,104]. An exclusively herbivorous diet is also known for M. glareolus in various regions [105].  The diet of M. arvalis and A. sylvaticus consists of a considerable but varying amount of invertebrates [106,107] explaining the higher variability of 4.0 and 5.3%, respectively. Mean values of these species correspond to a typical C 3 diet. The highest variabilities in d 13 C within the individuals from a single population were observed in R. norvegicus and M. musculus, with 10.2 and 12.7%, respectively. Both are synanthropic species with a very opportunistic diet that commonly includes human food supplies and food waste [108][109][110]. Thus, the high variability in d 13 C is likely caused by the presence or absence of C 4 -sugars and/or products containing tissues from animals fed on a C 4 diet in the rodent diet.

Variations in d 18 O CO3 and d 13 C between Incisor and Bone Material
The d 18 O CO3 values of the incisors differ by up to ,4% from that of bone material (Table S1), which can be explained by different time intervals in the mineralisation of both tissues. Rodent incisors (and also the molars of most arvicolids) are permanently growing and reflect a time span of a few weeks [86,87], whereas no pronounced remodeling of bone tissue occurs in rodents. Thus, bones archive a nearly life-long record [88]. This may also explain the phenomenon of why nearly 90% of the analysed individuals have higher d 18 O CO3 values in the incisors than in the bone material (Fig. 4). Taking into account the general population dynamics of small rodents in temperate regions with population maxima in late summer (see also materials and methods section), the incisor d 18 O CO3 of a large number of individuals are likely biased towards the 18 O-enriched precipitation of the warm summer months (Fig. 2). In contrast, bone material was formed over a longer time span, including the colder spring season or even the winter months in case of adult individuals that survived this period. This is further supported by the standard deviations of the bone-incisor offsets in each species (0.8 to 1.8%), which reflects the different times of mineralisation of both tissues in relation to a variable age structure of the individual specimens. One exception is the juvenile R. norvegicus specimens that have a relatively uniform bone-incisor offset within the five analysed individuals (,3%), with a small standard deviation of only 0.3%. This likely reflects their more or less identical age and hence similar time intervals recorded in bone and incisor d 18 O CO3 values.
The d 13 C values of the bulk incisor teeth, with two exceptions (Table S1), deviate less than 2% from the d 13 C results of the respective bone material, and no general trend to a skewed variation, as is the case for d 18 O CO3 , can be observed (Fig. 5). These variations can be explained by variations in the carbon isotope composition of the consumed food during the different mineralisation times of incisor and bone material.

Intra-jaw Variations in d 18 O CO3 and d 13 C of A. terrestris
The d 18 O CO3 of the molars of A. terrestris deviates from the incisor oxygen isotope composition of the same individual between 0.3 and 1.9%, indicating slight differences in the time periods of mineralisation of both tooth types. For other arvicolids (Microtus) a complete renewal of incisors within four to seven weeks is reported [87], whereas first and second molars of Microtus underly a complete renewal in a time interval of eight to twelve weeks [86]. Considering this, as well as the fact that the molars mostly have lower d 18 O CO3 values than incisors (Table S1; analogous to the incisor-bone offset), a small bias by seasonally driven population dynamics is likely the cause for the inter-tooth offset.
The difference in d 13 C between molar teeth and incisors of all analysed individuals is less than 2%, comparable to the boneincisor variations and reflecting varying carbon isotope composi-tions of the ingested food (Table S1). Given that the owl pellets represent a seasonally biased taphocoenosis, the lower d 13 C values in nearly all incisors compared to the molars of the same individuals can result either from a seasonal change in availability of food plants or from seasonal changes in the carbon isotope composition of the same plant species, which can be in the range of several per mil within one plant species (e.g., [67,111]).
Variations among M1 to M3 of the same animal are generally lower, both in d 18 O CO3 (0.2 to 1%) as well as in d 13 C (0.1 to 0.6%), indicating a more or less contemporaneous growth and mineralisation of these molars.
The D 18 O CO3-PO4 in A. terrestris Comparing incisors, molars and bone, the differences between d 18 O CO3 and d 18 O PO4 of each element overlap within error ( Table 3). The overall D 18 O CO3-PO4 of 10.960.8% is high and at the upper limit of previously reported values for large and medium-sized mammals (7.5 to 10.6%; [43][44][45]47,49,112]) that include perissodactyls, artiodactyls and carnivorans. The only reported data for rodents (laboratory rats, Rattus norvegicus) from two different settings show a highly variable D 18 O CO3-PO4 of 8.5 and 11.4%, respectively [48]. Further studies are needed to evaluate a possible species or body temperature dependence as suggested by Martin et al. [45] or a connection to seasonal variations in the d 18 O of drinking water and/or a food source of phosphate as supposed by Kirsanow and Tuross [48].

Minimum Sample Size Calculations
The bootstrapping results of our d 18 O CO3 data from rodents (more than 95% of the subsamples are within 61s of the mean if n sub $4, and even more than 99% if n sub $7) agree to the minimum sample size recommendations for most large mammals given in previous studies [28,29,32] as well as for the horses from the study of Wang et al. [30]. For goats and yaks, a considerably higher minimum sample size for reliable estimates of the population mean d 18 O within a 95% confidence level has been determined [30], presumably related to differences in physiology and/or a different nutritional or rather drinking behaviour. Our data for rodents indicate that the minimum sample size recommendations, determined for large mammals, potentially can be extended to small mammal species (i.e., rodents). Further taxon-specific studies, however, are needed to confirm if the present finding is valid for small mammals in general.
The regression line of Luz and Kolodny [4] is based on bone material from laboratory rats, and that of Longinelli et al. [36] is based nearly exclusively on bones of wild murids and arvicolids (including the genera Apodemus, Arvicola, Microtus and Pitymus). The regression of Navarro et al. [16] was primarily derived using molar and incisor teeth from wild small arvicolids of the genera Myodes and Microtus.
As the calibration equation of Luz & Kolodny [4] covers only a d 18 O LW range outside the d 18 O of local precipitation in our sample area, a direct comparison to our data is precarious. However, extrapolation of the regression line reveals that the bone-derived d 18 O values from the present study are much closer to this line than the tooth values ( Fig. 9).
A much better fit is obtained if the bone-based equation of Longinelli et al. [36] is compared with our bone data, which directly plot on the respective calibration line. A similar concordance exists by comparing our tooth-data with the toothbased equation of Navarro et al. [16] (Fig. 9).
This indicates that the choice of the skeletal element used for the calibration between the d 18 O of biogenic apatite and d 18 O LW has an important influence, which can be assigned to different mineralisation periods and seasonally driven population dynamics in rodents as discussed above. In the case of the study by Luz and Kolodny [4], an additional explanation for differences from our data could be the fact that the respective equation is based on laboratory animals that often differ in their physiology from field populations, which can have an impact on the oxygen isotope composition of their biogenic apatite (e.g., [82]). Furthermore, as laboratory rats were raised under controlled conditions with a constant diet and water source, this precludes the effect of any seasonal bias.
Previous studies have shown that tooth enamel should be the material of choice for oxygen isotope studies in fossil mammals, because it is less prone to diagenetic alteration than dentine or bone material [6,113]. Therefore, future d 18 O PO4 -d 18 O LW calibration equations of modern taxa that aim to reconstruct palaeoenvironmental conditions should be based on tooth material instead of bone. Furthermore, the same tooth type available also from the fossil taxa should be used, in order to avoid seasonal bias due to different tooth formation patterns.

Conclusions
The observed inter-specific variations in d 18 O CO3 within arvicolid and murid incisors are low. Intra-specific variations in the d 18 O of the investigated rodent bioapatite samples are in the same range as most previously observed values from large mammals. Statistical evaluations show that minimum sample size recommendations from existing studies on large mammals agree well with the results from the present study, indicating at least 95% confidence that from a subsample size of $4 mean values range   10.960.8% determined for A. terrestris is at the upper limit of data from large mammals and ranges within the two previously reported values from another rodent species (R. norvegicus). Within the present study, no significant disadvantages in the interpretation of d 18 O data from rodent bioapatite compared to large-and medium-sized mammals were observed. Nevertheless, it has to be taken into account that fossil rodent taphocoenoses in temperate regions mostly underlie seasonal and spatial predator-prey population dynamics.
The d 13 C data mirror the different nutritional strategies of the analysed species (i.e., dietary specialisation vs. dietary opportunism). The range of variations within a single species can vary considerably but agrees well with field observations of the respective nutritional behaviour. Thus, multiple analyses from different individuals are an important prerequisite to get deeper insight into specific nutritional adaptions from fossil small mammals. Seasonally driven variations in the carbon isotope composition of the ingested food are reflected in different skeletal tissues, caused by differing time intervals of incisor, molar and bone mineralisation.
The present study underscores the high potential of stable isotope signatures of fossil small mammal skeletal remains to reconstruct past environmental and climatic conditions. However, further detailed systematic investigations of modern analogous communities of small mammals are necessary for a refined understanding of the incorporation of oxygen and carbon isotope compositions in the bioapatite of their teeth and bones and to explore the full potential of fossil rodents for palaeoenvironmental research.

Supporting Information
Table S1 Detailed summary of analytical results (XLS)