Study design

We measured δ¹³C values in lower first molars (m1) of murine fossils collected from the Siwalik Group in the Potwar Plateau of northern Pakistan, ranging from13.8 to 6.5 Ma (21 localities), and northern India, ranging from ∼2.5 to ∼1.8 Ma (2 localities), and compared them with the compiled data of large herbivorous mammals and soil carbonates presented in Figure 2 of Badgley et al. [23]. Carbon isotopes in enamel reflect dietary input during the time of tooth formation, whereas dental morphological characters are long-term adaptive traits obtained through evolutionary history. Two ecomorphological characters, van Dam's [30] index (VD index hereafter, defined below) and tooth crown height (hypsodonty) were evaluated to compare timing and direction of morphological change with isotopic dietary inferences.

To minimize inter-tooth isotopic variation, we deliberately analyzed only m1 (except several samples from the Pliocene of India). Isotope compositions across teeth in different tooth positions are influenced by nursing to different degrees (Figure S1) because mineralization of molar enamel starts before weaning in murine rodents (see Supporting Information S1). Although we do not exactly know the carbon isotope effect of nursing on murine molars, the magnitude of relative change across the data of m1 should reflect actual changes in the Siwalik murines, assuming that physiological differences in the closely-related species are negligible. We chose m1 over the second and third molars even though m1 is most influenced by mother's milk because m1 is large enough (1.5 to 2.9 mm in length) to generate a CO₂ sample from a single tooth, and morphological complexity of m1 is useful for systematic identification and allows for species-level comparisons.

Materials

Samples ranging in age from 13.8 to 6.5 Ma were collected by sieving in the 1970's to 2000 from the Potwar Plateau, northern Pakistan. The fossil specimens are on long-term loan from Pakistan, under the authority of the Geological Survey of Pakistan, Islamabad, Pakistan, and are housed in the Peabody Museum of Archaeology and Ethnology, Harvard University. To fill the time gap between 6.5 Ma and Recent, five specimens from Kanthro (∼2.5 Ma) and six specimens from Nadah (∼2.0 to ∼1.8 Ma) recovered from Upper Siwalik localities of northern India [31], [32] were analyzed. All Indian specimens used in this study are housed in the Centre of Advanced Study in Geology, Panjab University. Recent specimens from both regions were also utilized for comparison. Among those from Pakistan, Golunda ellioti and two Mus species (M. booduga, M. saxicola) were from owl pellets collected on the Potwar Pleatau, whereas Rattus sp. and Millardia sp. were captured in urban areas of the Potwar between 1970 and 1975. They are housed in the Shuler Museum of Paleontology, Southern Methodist University. Bandicota indica was collected from an owl pellet in Nagar, Punjab State, India in 2011. Specimen numbers are given in Datasets S1 and S2. No additional permits were required for the described study, which complied with all relevant regulations.

Specimen Identification and Sample Selection

Jacobs [27], [33] described rodent specimens from localities YGSP 491 (13.8 Ma), YGSP 41 and 430 (13.6 Ma), YGSP 182 (9.2 Ma), DP 13 (6.5 Ma), and DP 24 (∼1.7 Ma), and recognized 9 species among 8 genera. Cheema et al. [34] described Progonomys hussaini from locality JAL-101 (∼11 Ma). Specimens from other localities have been identified at the generic level or, if not, were assigned to ambiguous groups such as “near Antemus” according to the most updated study by Jacobs and Flynn [29]. The qualitative nature of the taxonomic assignments and anagenetic change in time-consecutive taxa impede the systematic classification of the Siwalik murines. We grouped tooth samples at each stratigraphic level based on size and morphology, so that individuals in the same assemblage display high similarity and generally followed Jacobs and Flynn [29] for names of taxa (Table S1). We analyzed all murine species known with n>5 by upper first molar (M1) at each stratigraphic level, as well as large Karnimata sp., which are minor components (n<5 by M1) at 8.2 and 8.8 Ma (Table S1). Complete m1 with slight to moderate wear, corresponding to wear stage I to IV of Lazzari et al. [35], were selected for carbon isotope analysis. The tooth samples are neither weathered nor etched. The enamel is intact based on visual observation under a light microscope with 100× magnification for all and SEM observation for randomly selected specimens. Diagenetic alteration was assumed to be negligible for the murine rodent teeth as confirmed by Quade et al. [36] for large mammal teeth from the Siwalik sediments of similar ages in the same region.

Carbon Isotope Analysis

We conducted in-situ laser ablation GC-IRMS isotope analyses on enamel of m1 using the laser ablation method described by Passey and Cerling [17] at the University of Utah. Isotope ratios are expressed as δ values (‰) on VPDB scale, where δ_sample  =  (R_sample/R_standard-1)*10³, and R is given by ¹³C/¹²C. Fossil specimens were cleaned by acetone to remove glue and were treated with 2% NaOCl for one hour to remove organic contaminants, followed by 0.1 M sodium acetate-acetic acid buffer for one hour (except specimens from YGSP 182, see below) to remove diagenetic carbonate minerals from enamel surfaces. The time period for the 2% NaOCl treatment varied from 3 hours to 8 hours in modern specimens, depending on the amount of organic tissue adhered to the specimen. The 0.1 M acetic acid buffer (pH = ∼5.4) was used rather than 0.1 M acetic acid (pH = ∼3.0) because thin enamel of murine molars are etched by 0.1 M acetic acid in only 15 minutes as observed under SEM (Figure S2). The treated samples were mounted on a plate of the laser sample chamber using Bostik Blu-Tack. The lingual side was preferentially used, but the labial and posterior sides were also analyzed when data results from the lingual side were not reliable due to low CO₂ yield or contamination of CO₂ generated from ablation of dentine or Blu-Tack. CO₂ laser (wave length: 10.6 μm) setting ranged from 1.8 to 7.5 W with a 8.5 ms pulse duration. CO₂ from multiple ablation pits was cryogenically concentrated, inlet to a gas chromatographic column (Poraplot Q; 60°C) and sent into a Finnigan MAT 252 gas-ratio mass spectrometer via a GC/CP interface.

Isotope data reported in this study were corrected for isotope fractionation between laser and conventional H₃PO₄ methods to be comparable with the compiled data of Badgley et al. [23], in which carbon isotope values were obtained by the H₃PO₄ method. Isotope enrichment (ε) between two substances A and B is defined as ε_A−B  =  (R_A/R_B −1)*1000 [2], [37], which is independent on the scale (i.e., PDB, SMOW) and more preferable than the scale-specific difference of Δ_A−B  =  δ_A–δ_B [2], [38]. The superscript * (ε*) is designated for non-equilibrium processes [2]. The mean ¹³ε*_laser-H3PO4 value was −1.5±0.3‰ (1σ) for a modern beaver incisor analyzed in every run, whose isotope values had been determined by the H₃PO₄ method. This value is larger than the value (¹³ε*_laser-H3PO4 = −0.3‰) reported by Passey and Cerling [17] but is smaller than those of Podlesak et al. [39]. The mean ¹⁸ε*_laser-H3PO4 value, −6.1±0.9‰, was a much larger offset than that of ¹³ε*_laser-H3PO4 values. This is because laser ablation of enamel liberates both phosphate and carbonate bound oxygen, whereas the H₃PO₄ method targets only carbonate oxygen. There is a ∼9 permil offset between phosphate and carbonate bound oxygen [40]. Oxygen isotope data produced by the CO₂ laser ablation method are less precise than isotopic measurements of phosphate oxygen or structural carbonate oxygen due to incomplete mixing of oxygen-bearing components [17]. We do not discuss oxygen isotope compositions of Siwalik murine rodents in this study. Carbon isotope compositions of Recent species reflect light carbon inputs to the atmosphere by burning fossil fuels. They were corrected for δ¹³C values (−6.5‰) of atmospheric CO₂ in pre-industrial times [41]. Modern δ¹³C _atmosphere in years of sampling (1970's and 2011) are −7.4‰ [42] and −8.2‰ [43], respectively.

Nine specimens from YGSP 182 (Dataset S1) were treated with 0.1 M acetic acid for 8 hours and resulted in strongly etched enamel surface. These initial treatment results led us to check the effect of the acetic acid treatment on murine enamel. By observation under SEM, removal of non-prismatic outer enamel started in only 15 minutes in a 0.1 M solution due to the low pH of the acetic acid (pH = ∼3.0) (Figure S2). Thus, the 0.1 M acetic acid treatment was replaced with 0.1 M sodium acetate-acetic acid buffer (pH = ∼5.4). The isotopic composition of the YGSP 182 specimens may have shifted by the acetic acid treatment. In Koch et al. [44], powdered enamel samples treated with 0.1 M acetic acid for three days were altered by −0.2 ‰ for δ¹³C and by +0.9 ‰ for δ¹⁸O compared to untreated samples. In the YGSP 182 specimens, the mean isotope ratio of the etched samples (n = 9) are offset by −1.1‰ in carbon and by +1.5‰ in oxygen, compared to samples (n = 2) treated by buffered acetic acid. Although the direction of the acid effect is the same as demonstrated in Koch et al. [44], the absolute values for correction are not known in this study. The nine specimens are not corrected for the difference in the acid treatment.

Quantifying Dental Characters

The VD index evaluates space between anteroposteriorly aligned cusps on upper first molar (M1), defined as the ratio of tooth width to a distance between the posterior side of the lingual anterocone and that of protocone. Species with a value of more than 2.2 are predominantly grazers [30], [45]. Note that van Dam [30] used two other characters along with VD index in a principle component analysis in order to distinguish grazing murines from non-grazers. Both characters are based on m2, which are more difficult to identify and are not used in this study. Following the result of van Dam [30], taking 2.2 as a critical value for grass diets is reasonable although it does not necessarily mean that every species with >2.2 is a grazer.

Hypsodonty is well-studied in ungulates and is associated with adaptation to high rates of tooth wear due to tough diets together with ingestion of soil and grit while feeding [46], [47]. Hypsodonty index is most commonly determined by height/width of unworn m3 [48]. In this study, hypsodonty was measured as crown height divided by length of M1. The length was used instead of width because its variance was less than half that of width among the species identified by Jacobs [27], [33].

Specimens used for VD index are unworn to moderately worn, corresponding to wear stage I to IV of Lazzari et al. [35], and those used for hypsodonty are unworn or slightly worn specimens in wear stage I to II. All dental measurements were taken in VHX-1000 communication software based on 2D digital pictures photographed with a Keyence VHX-1000 digital microscope. In a mean of VD index for each species, large Karnimata sp. at 9.2 Ma (YGSP 7717) was included in K. darwini due to small sample size.

Statistical Analysis

The rates of temporal change in δ¹³C values were compared between two anagenetic lineages, Karnimata of the Karnimata clade (blue circles in figures) and the Progonomys clade without Mus sp. at 7.4 Ma (red triangles in figures), from 9.2 through 6.5 Ma by using linear regression lines between δ¹³C values (dependent variable) and time (independent variable, covariate) of the two clades (categorical variable). That species of Mus sp. was excluded because it may be an immigrant from another region, based on its smaller size and more elongated anterior side of the lingual and labial anterocones than those of the younger Mus auctor (6.5 Ma). To avoid confusion, the Progonomys clade without Mus sp. at 7.4 Ma is expressed as the “Progonomys clade” hereafter. We tested the interaction between the categorical variable and the covariate in the aov() command of R for a null hypothesis that they do not interact. Because the null hypothesis was accepted, i.e., slopes of the two regression lines are equal (p = 0.54), Analysis of covariance (ANCOVA) was performed to test if intercepts of the linear regression lines between the two clades are significant. Before performing the linear regression, we checked that a linear model fitted the data as well as or better than a quadratic model. Temporal changes in VD index and hypsodonty values between the two clades were also examined in the same procedure. We did not adopt linear regression analysis between the ecomorphological characters and δ¹³C values (e.g., VD index vs. mean δ¹³C values) because the variables are at different time scales. Carbon isotope compositions in enamel are a short-term variable, reflecting mixing of two end-members, C₃ plants and C₄ plants, in their diets during enamel mineralization. On the other hand, VD index and hypsodonty are long-term variables that species acquired along phylogeny. As a result, linear regressions between the ecomorphological characters and δ¹³C values have lower R² values than linear regressions of temporal changes in the three variables.

At 7.4 Ma, 6.5 Ma, and Recent, in which three species were considered, Welch's ANOVA was performed to assess whether means of δ¹³C values in the three coexisting species are indistinguishable at α = 0.05. Welch's ANOVA was chosen to avoid increasing Type II error rates due to small sample sizes. The Games-Howell test was used for post-hoc tests to identify the location of significant differences at α = 0.05.The assumptions of normality and homogeneity of variance were checked by the Shapiro-Wilk test and the Bartlett's test. Recent Millardia sp. was combined with Recent Rattus sp. due to small sample size. Although molecular studies show they are not very closely related [49], Millardia was sometimes considered a subgenus of Rattus [50]. Recent Mus spp. includes Mus booduga and Mus saxicola. The 95% bootstrap confidence intervals were computed by the bias-corrected and accelerated method with 9999 randomizations in the ‘boot’ package [51]. All statistical tests but Welch's ANOVA and Games-Howell test, which were computed in IBM SPSS Statistics 18 [52], were performed in R2.15.1 [53]. Samples whose values clearly deviate from others in the same species and are beyond three standard deviations were removed from statistical tests as outliers.

Comparisons with Carbon Isotope Ratios in Large Mammals and Soil Carbonates

An abrupt positive shift of δ¹³C values indicates a pronounced dietary shift of murine rodents between 7.8 and 7.4 Ma. The dietary shift is more abrupt in murine rodents (Figure 1) than in large mammals [23]. Comparisons at different systematic levels may contribute to the fact that large mammals have wider variation of δ¹³C values than murine rodents. However, the ranges of δ¹³C values in murine rodents are concordant with variation within 5 to 95 percentiles of δ¹³C values in large mammals, ∼3‰ for 13 to 11 Ma and ∼9‰ for 7.4 to 7.0 Ma. Hipparionine equids incorporated C₄ grasses into their diet by ∼8.5 Ma [19], [23], whereas murine rodents lagged behind these ungulates by one million years. The pattern of δ¹³C values in murine rodents is similar to the isotopic shift recorded in soil carbonates, which reflect the overlying vegetation more directly than large mammal data. It suggests that murine rodents record past ecological conditions more precisely than large mammals because of their feeding behavior as dietary opportunists with small home ranges in short life span.

In contrast, neither large mammals nor soil carbonates record the gradual negative shift of 2.5‰ toward 8.7 Ma observed in murine data. The negative shift may reflect an actual change in vegetation because the most negative mean δ¹³C values in large mammals and soil carbonates are at 9.2 to 8.6 Ma and at 9.2 to 8.3 Ma, respectively, which encompasses the 8.7 Ma level of the murine data.

Dietary Niche Partitioning

From 9.2 Ma through 6.5 Ma, large murine species have a consistently greater mean δ¹³C value than small species with the exception of Parapelomys robertsi (Figure 2). Carbon isotope ratios in coexisting species can vary due to different patterns in spatial occupation. In a modern ecosystem, Codron et al. [54] showed δ¹³C values of C₃ plants vary by up to 2‰ in an African savanna, having more negative values associated with availability of perennial water sources. Thus, the isotopic variation (less than 1.7‰) between large and small murine species observed from 9.2 Ma through 8.2 Ma may be explained by spatial partitioning, with small species feeding on less water-stressed C₃ plants. Considering isotopic enrichment of insects and non-photosynthetic plant tissues relative to green leaves [55], [56], the magnitude of the isotopic variation can also be explained by dietary niche partitioning, with large species feeding on more strictly vegetarian diets. The intake of C₄ grasses might be partially responsible for the isotopic differentiation as early as 8.2 Ma because isotopes in hipparionine equids indicate a C₄ dietary component beginning at 8.5 Ma in Pakistan [23]. Nevertheless, C₃ vegetation still dominated major floodplains at ∼8.0 Ma [57].

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Figure 2. Scatter plots of δ¹³C values vs. the natural logarithm of tooth size, Ln(length*width), of m1 in murine rodents, ranging from 9.2 Ma to 6.5 Ma and Recent.

Outliers removed from statistical tests are shown by asterisks.

https://doi.org/10.1371/journal.pone.0069308.g002

The isotopic variation of less than 2‰ may also result from proportional contributions of insects and non-photosynthetic tissues of C₃ plants in the diet of murine rodents. Most of today's murine rodents are omnivorous [50], consuming a variety of plant material (seeds, grains, nuts, fruits, leaves, stems, roots), insects, and other invertebrates. Insects tend to be more enriched in ¹³C than primary producers due to trophic enrichment [55]. The mean estimate of trophic shift is +0.6‰ among terrestrial arthropods, which are one of major food source for murines, based on data compiled in McCutchan et al. [55]. The isotopic differentiation resulting from different proportional amounts of insects is observed in modern Rattus rattus and Mus musculus inhabiting forests of Hawaii, where they partition food resources in that Rattus rattus is primarily vegetarian (∼80% of stomach content being fruits and seeds), whereas Mus musculus consumes a large amount of arthropods (∼50% arthropods, ∼30% fruits and seeds) [58]. A mean δ¹³C value in bone collagen of Rattus rattus is more negative than the more insectivorous Mus musculus by ∼1.4‰ [58]. Non-photosynthetic plant tissues are slightly more enriched in ¹³C than leaves [59]. Badeck et al. [56] documented that roots and woody stems are isotopically heavier than leaves by 1.1‰ and 1.9‰ on average in C₃ plants, respectively, whereas the differences between roots and leaves are not significant in C₄ plants. Fruits and seeds are also more enriched than leaves. The isotopic differences between fruits and leaves (Δ¹³C _{fruits-leaves}) range from 0.5‰ to 3.0‰ in various species of leguminous plants [60], [61].

By 7.4 Ma, the difference in mean δ¹³C values between large (Karnimata sp.) and small (Mus sp.) species increased significantly, reaching 3.5‰ (Figure 2 and Table S6). The difference of 3.5‰ is too great to be explained solely by isotopic variation in C₃ plants or physiological variation in murine species and therefore indicates that consumption of C₄ grasses had come to play an important role in isotopic differentiation among coexisting species. If the difference arose solely from proportional contribution of C₄ to C₃ diets, we conclude that Karnimata sp. preferentially consumed C₄ plants over C₃ green leaves 20 to 30% more than Mus sp. at 7.4 Ma. The isotopic difference of 4.3‰ at 6.5 Ma, between Parapelomys robertsi and Karnimata huxleyi, translates into 30 to 40% more consumption of C₄ grasses by K. huxleyi.

Comparisons with Dental Morphology

The tooth size of each individual is expressed as a natural logarithm of tooth area, Ln(length*width), in Figures 2 and 3A, B (Table S7). Two sympatric species (or morphotypes) can be recognized as early as 11.2 Ma based on dental morphology of M1 [29]. These species greatly overlap in size and morphology of m1. They are tightly clustered by carbon isotope composition (Figure 1, Table 1), indicating that these species of similar size did not isotopically partition their diets. This condition lasted at least until 10.1 Ma. Size divergence occurred by 9.2 Ma (Figure 3A, B). Compared to corresponding species at 8.2 Ma, by 7.4 Ma, Progonomys sp. became larger, while Karnimata sp. reduced its size, resulting in opening morphospace for small Mus sp. and large Parapelomys sp. The species of Mus sp. was a major species at 7.4 Ma (Figure 3A). The presence of three major species (P. robertsi, K. huxleyi, M. auctor) continued through 6.5 Ma (Table S1).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Figure 3. Change in morphological characters through time.

(A) Ln(length*width of m1). Symbols in A: 13.8 to 13 Ma, inverted orange triangle for Antemus chinjiensis; 11.6 to 11.2 Ma, purple diamond for Progonomys hussaini +?Karnimata sp.; 10.5 to 10.1 Ma, blue cross for Karnimata sp. + Progonomys sp.; 8.8 Ma, green cross for morphotype 7; 7.8 Ma, green cross for morphotype 8; 7.4 Ma, light blue square for Parapelomys sp.; other symbols as in Figure 2. (B) 20th percentile and 80th percentile of Ln(length*width of m1). (C) van Dam's [30] index of M1, calculated as tooth width divided by the distance (lap in figure) between the posterior side of the lingual anterocone and that of protocone. The vertical bar at 2.2 shows the lower boundary of VD index to be predominantly grazers. Symbols in C: 12.4 Ma, purple diamond for near Progonomys sp.; 11.6 to 11.2 Ma, purple diamond for Progonomys hussaini; 11.2 Ma, blue circle for ?Karnimata sp.; 10.5 and 10.1 Ma, blue circle for Karnimata sp., red triangle for Progonomys sp.; 8.8 and 8.2 Ma, blue circle for Karnimata sp. (+ large Karnimata sp.); other symbols as in A. (D) Hypsodonty of M1, calculated as crown height divided by length. Symbols as in C. Error bars indicate 95% bootstrap confidence intervals. Red dotted lines connect two reference points, and blue dotted lines are parallel to the red line and have one reference point.

https://doi.org/10.1371/journal.pone.0069308.g003

Among coexisting species that differ in size, large species consistently have greater VD index and higher hypsodonty values (i.e., narrower valleys between anteroposteriorly aligned cusps on higher-crowned teeth) than small species from 9.2 through 6.5 Ma (Figures 3C, D, Tables S8 and S9). Karnimata sp. progressively narrowed valleys between anteroposteriorly aligned cusps through the evolutionary sequence, indicating that they consumed greater amounts of tough diets through time. In the VD index, Karnimata sp. exceeds the grazing value of 2.2 by 7.4 Ma, and large species (Parapelomys robertsi and Karnimata huxleyi) are above that value at 6.5 Ma (Figure 3C). In Karnimata, the attainment of the VD value signaling consumption of grass diets corresponds in time to the increase of the mean δ¹³C value and increased total range of δ¹³C values at 7.4 Ma. On the other hand, the “Progonomys clade” did not modify the valley space between cusps as shown by a constant VD index value even though the rate of dietary change toward more tough diets is same as Karnimata. At 7.4 Ma, Mus sp. has a mean VD value (1.77) significantly lower than that of Progonomys sp (1.96). Although the rates of change in δ¹³C values are the same in both clades, the different rates of morphological evolution in VD index indicate that selection pressures leading to the morphological change are greater in Karnimata than in the “Progonomys clade”.

There is no evidence of increasing crown height in Karnimata and the “Progonomys clade”, but Parapelomys robertsi, which is a derived member of the Karnimata clade [28], has significantly higher tooth crowns than the coexisting species at 6.5 Ma (p≤0.01 for each pair in the Games-Howell test). Thus, in Siwalik murines, tooth crown height lagged attainment of the grazing threshold in the VD index by 2.7 Ma. A morphological trend toward increasing VD index and hypsodonty values was also recognized in a clade of European murine rodents, Progonomys-Occitanomys-Stephanomys [30].

Contradictions in Carbon Isotope and Dental Morphology of P. robertsi

At 6.5 Ma, the largest species, Parapelomys robertsi, is more depleted in ¹³C than its two coexisting species (Mus auctor, Karnimata huxleyi), which would indicate a higher proportion of C₃ plants in the diet. However, P. robertsi is expected to be a grazer based on dental characters. Its VD index value (2.5) is equivalent to that of Recent Golunda ellioti (Table S8), which predominantly consumes C₄ grass. Its tooth crown is higher than any other Siwalik murines (Figure 3D and Table S9). Similar apparent contradictions in dental morphology and isotopic indicators of diet can be found in modern rodents. Recent Millardia sp. shows more negative δ¹³C values than Recent Mus booduga and Mus saxicola, but dental characters of Millardia sp. appear to be more adapted to a grass diet than for the two species of Mus (Tables S8 and S9). Shiels [58] reported that Rattus rattus has more negative δ¹³C values than coexisting Mus musculus although Rattus has greater values in VD index than Mus (Table S8). Among large herbivorous mammals, hypsodont ungulate species are not universally specialized for consuming C₄ plants [6], [62]. Feranec [6], [62] proposed that hypsodonty may broaden the dietary niche of generalists feeding on mixed C₃/C₄ diets, which may be the case for murine rodents. Further studies of isotopic dietary inference are necessary to understand the ecomorphological strategy of small mammals. However, taxonomically fine-scale comparisons of carbon isotope ratios with grazing indices throughout the temporally well-constrained sequence of Siwalik fossil localities has provided a more detailed and calibrated view of rodent ecomorphology and dietary change than previously possible.