Variation in the stable isotopes of land-snail shells potentially provides useful information for reconstructing the terrestrial paleoenvironment. In this study, we investigated the relationship between vegetation and variation in the shell carbon (13C/12C) isotope ratios in the land snail Mandarina chichijimana on Chichijima of the Ogasawara Islands. In modern samples, the mean δ13C value of the adult shell (range -13.8∼-9.6) was higher at sites that contain a greater proportion of C4 plants, especially near the coast (regression equation, mean δ13C=6.594×[C4 plant proportion]-12.43). The shell δ13C of the living snails was also significantly positively correlated with the δ13C of the body tissues. As no differences were found in the δ13C values of shells collected in carbonate-rich areas and volcanic rock settings, we conclude that δ13C in M. chichijimana is almost exclusively influenced by the plants that they consume. Also, in fossil shells from limestone outcrops, the mean δ13C value (-11.6∼-11.4) was significantly lower than in modern shells from the same limestone outcrops (-11.0∼-9.6). This is therefore preliminary evidence that C4 vegetation declined in line with a decrease in sea level around the time of the Last Glacial Maxima. Together, the findings may form a basis for the future use of land snail shells to estimate the paleoenvironment of the Pacific Islands in this region.
Introduction
The composition of δ13C in land snail shell carbonate is influenced primarily by diet, so can be used to estimate the paleoenvironment (Goodfriend and Hood, 1983; Goodfriend, 1999; Baldini et al., 2007; Colonese et al., 2007; Yanes et al., 2008). This depends on the fact that metabolic CO2 derived from the diet (released as respiratory CO2, which dissolves in the body fluids) contributes to shell δ13C through shell formation. However, a number of other environmental and physiological factors (e.g., ingestion of carbonates, exchange with atmospheric or soil CO2, and metabolic rate) may also affect shell δ13C (Yanes et al., 2008). There is therefore some debate as to the extent that this limits the use of snails in general in paleoenviromental reconstruction (Stott, 2002; Metref et al., 2003; Balakrishnan et al., 2005).
Certainly, some studies have shown a clear effect of diet on shell δ13C, with little or no influence from ingested CaCO3 (Stott, 2002; Metref et al., 2003). Generally, 13C tends to become enriched in the shell carbonate of snails with a diet of C4 plants (e.g., grasses) relative to those that have a diet of C3 plants (e.g., trees) (Goodfriend and Ellis, 2002; Stott, 2002; Metref et al., 2003; Balakrishnan and Yapp, 2004; Baldini et al., 2007; Yanes et al., 2008). This knowledge should in principle enable the δ13C values of fossil shells to be used to infer the paleoenvironment (Goodfriend, 1992). However, as several assumptions about the feeding behavior and physiology are still required, e.g., snails consume C3 and C4 plants indiscriminately and in proportion to abundance of these plants (Goodfriend and Magaritz, 1987; Goodfriend, 1988), it is first essential to develop modern isotopic references, by exploring the relationship between carbon isotopic composition of modern snail shells and present-day habitats. Accordingly, the endemic land snail Mandarina of the Ogasawara Islands may be useful to address these issues because of its limited dispersal ability, wide range of resource use, occurrence of abundant fossils and high variability in habitat condition (Chiba, 1996, 2004).
Mandarina have undergone an adaptive radiation within Ogasawara, so show diversification in habitat use, e.g., arboreal or ground-dwelling (Chiba, 1999; Davison and Chiba, 2006). Shell morphology is also known to be associated with the habitat (Chiba, 2004; Chiba and Davison, 2007). For example, snails inhabiting thick litter of palm or pandanus leaves have a shell that has a higher spire than those inhabiting thin litter of broad-leaved trees (Chiba and Davison 2007). As fossil records of Mandarina show that morphological changes have occurred during 40,000 years (Chiba, 1996, 1998), the suggestion is that these were partly caused by Pleistocene changes in vegetation. Vegetation may have itself changed because of cooler sea surface temperatures (SST) in the equatorial/tropical regions of the Western Pacific (Visser et al., 2003; Stott et al., 2002) and greater aridity, most likely due to an increase in the length of the dry season (e.g., Moss and Kershaw, 2000). Unfortunately, there is still much uncertainty in the paleoclimatic reconstructions, so novel methods of estimating the paleoenvironment of the region are useful.
The diet of Mandarina is fallen leaves for ground-dwelling species and still-live leaves for arboreal species, some of which originate from C3 or C4 plants. Thus, variation in the shell δ13C and morphology can be combined to clarify how the habitat changed during the Pleistocene. In addition, another benefit is that if the shell δ13C values may be used to estimate dietary preferences and habitat use among different species, then stable isotope analysis of the empty shells may be a noninvasive technique to investigate the native, endangered malacofauna.
In the present study, variation in shell δ13C was investigated among modern snails of Mandarina chichijimana Chiba, 1989 collected from different vegetation (C3 and C4) and environments (e.g., grassland near the beach, inland forest, limestone outcrops, volcanic rock areas) on the island of Chichijima. Relationships of shell δ13C with environmental factors are examined. In addition, change of the habitat since the end of the last glacial era is estimated through the carbon stable isotopic analysis of fossil M. chichijimana shells.
Materials and methods
Samples and habitat analysis
Modern shell samples of the ground-dwelling species, Mandarina chichijimana, were collected from 14 localities in the southern parts of Chichijima during 1987–1989 (Figure 1). At each locality, a 10×10 m quadrat was placed and fresh empty shells (with periostracum) were collected from the quadrat. Old empty shells were avoided because the habitat may have subsequently changed.
Each quadrat was subdivided into one hundred 1×1 m squares, and the plant species found in the litter within the 1×1 m squares were recorded, also separating them into C3 and C4 plants. The frequency of the C4 plants within each 10×10 m quadrat was obtained by calculating frequency of the 1×1 m squares in which the area covered by C4 plants was greater than that by C3 plants. To investigate variation in δ13C among the species of C3 or C4 plants, fallen leaves of the plant species that dominate each quadrat were collected to analyze 13C/12C ratios. The leaf samples were kept frozen between the time of collection and analysis.
Figure 1.
Maps of the Ogasawara Islands, Chichijima and southern parts of Chichijima with locations of sampling sites. Numbers in parentheses indicate size of the sample used for shell δ13C analysis.

The samples M9–M14 were all from the small limestone outcrop (Minamizaki) near the beach (Figure 1). All other samples were collected from volcanic rock areas. The samples M1 and M2 were also from the site near the beach, but other modern samples were all from inland forest. In addition to the modern empty shells, two live snails were collected from each of the three localities of the limestone outcrop (M10, M11, M12) and the three localities of volcanic rock area (M3, M4, M8). Except for these samples, living snails were not used in the present radiocarbon analysis, because the species is critically endangered.
The Pleistocene fossil samples of M. chichijimana (F15, F16) were collected from the fissure deposits on the Minamizaki limestone outcrop. Details of the fissure deposits and geological background of the fossils were described in Chiba, 1996. The ages of the fossil samples used in the present study were 18230 yr BP and 22300 yr BP. (Chiba, 1996).
Table 1.
Sample size (modern empty shells, fossil shells, and shells and soft parts of live individuals) and geographical coordinates, geology and habitat of the sample localities.

In total, shells of 111 modern empty snails, 12 live snails (6 from limestone outcrops and 6 from non-limestone areas) and 16 fossil snails and soft parts of 12 live snails were used for δ13C analyses (Table 1).
Stable isotope analysis
The fossil shells used for the analysis were all adults, because the shells with a thick outer lip of the aperture were used. The modern shells used for the analysis were young adults, because their shells had a slightly thickened outer lip of the aperture. The shell from the outer lip to the part approximately ⅛ whorl before the outer lip was used to determine δ13C for both modern and fossil samples, because habitat use may differ between juvenile and adult snails.
Shell was first soaked in diluted HCl to remove soil and calcium salt on the shell surface. After being washed in water, the shells were then soaked in distilled water, and dried at 80°C. A 10–15 mg aliquot was used for isotope analysis using the standard method by McCrea (1950). Snail soft parts and plant leaves were dehydrated in air at 50°C, homogenized and converted to powder. Each aliquot represented an entire-averaged specimen. The 13C/12C ratios of the released CO2 were determined using VG Optima (VG isotec) and Delta S (FinniganMat) mass spectrometer. The carbon isotope compositions are reported per mil (‰) relative to the Vienna Pee Dee Belemnite (VPDB) standard (Craig, 1957) as delta according to the following equation:
The precision of the shell δ13C in the present analyses was ±0.1‰, obtained by multiple measurements of international standard aliquots. These analyses were conducted at the Geoscience Laboratory Inc (Nagoya, Japan) and Institute of Accelerator Analysis Ltd (Kawasaki, Japan).Statistical analysis
To examine possible associations of shell δ13C with the frequency of C4 plants in the litter at each sample locality, we used a generalized linear mixed model (GLMM) to control for unobserved heterogeneity across individual snails. The GLMM included locality as a random effect, and geology (limestone or volcanic rock) and frequency of C4 plants in the litter in each 10×10 m quadrat as fixed effects. The analysis of the association between shell δ13C and C4 plants was conducted using all of the modern samples. The modern samples of the limestone outcrops were also analyzed separately to examine the consistency of the relationship between shell δ13C and C4 plants over different spatial scales. In addition, a regression analysis was conducted between sample means of shell δ13C and frequency of C4 plants at the sample locality to obtain a simple quantitative relationship between the two variables. The relationship of δ13C between shell and soft parts was also examined for live snails by regression analysis. The GLMM was estimated using maximum likelihood in R 2.6.1 (R Development Core Team, 2006) and the ‘nlme’ library (Pinheiro and Bates, 2000).
Table 2.
Sample means and standard deviations of δ13C obtained from modern empty shells and fossil shells. Individual δ13C values obtained from shells and body tissue of the live individuals are also shown.

Results
It was found that in each quadrat, either C3 or C4 plants dominated. The δ13C values of these plants ranged from -32.9 to -25.5‰ and from -13.5 to -13.3‰, respectively (Table 1). The C4 plants are restricted to the localities near the beach (M1–M2, M9–M14), and no C4 plants were found in the inland areas (M3–M8) (Figure 2).
The shell δ13C was highly variable, ranging from -15.7 to -4.6‰. In modern samples, the sample mean of the shell δ13C value ranged from -13.8 to -9.6 (Table 2) at each locality. Large variations in the shell δ13C were found not only among the samples (Table 2, Figure 2), but also within the samples (Table 2, error bars in Figure 2). When all data were pooled, the δ13C values of modern land snail shells were positively correlated with the frequency of C4 plants in the 10×10 m quadrat (F(1,10) =62.28, P<0.001). No differences were found in the δ13C values between samples from the limestone outcrops and those from the volcanic rock areas (F(1,10)=1.37, P=0.269) (Table 3, Figure 2). The relationship between sample mean of shell δ13C and frequency of C4 plants in the quadrat was evidenced by a regression equation: mean δ13Cshell=6.594×(C4 plant proportion)—12.43.
Figure 2.
Scatter plots between sample means of the shell δ13C and frequency of the C4 plants of the sampling sites. The error bars indicate two standard deviations. The mean shell δ13C of the fossil samples are also shown. Closed circles: samples from the limestone outcrop, closed triangles: samples from near the beach of the volcanic rock area, open circles: samples from inland of the volcanic rock area, closed square: fossil samples.

In the analysis using only modern samples from the Minamizaki limestone outcrop, the shell δ13C was significantly positively correlated with frequency of C4 plants in the quadrat (b=4.074, F(1,4)=10.924, P=0.029) (Figure 2). The shell δ13C was significantly higher in the modern samples from the Minamizaki than in the fossil samples from the same area (F(1,6) =7.724, P=0.032) (Figure 2).
The δ13C of the soft parts obtained from 12 living snails ranged from -25.0 to -23.0. The shell δ13C was significantly positively correlated with δ13C of the soft parts (R=0.721, P=0.008) (Figure 3). The calculated regression equation was: δ13Cshell=0.912×δ13Csoft part+9.815. The slope of the regression line was 0.912, and the 95% confidence interval of the slope ranged from 0.712 to 1.116.
Discussion
The present study shows that the shell δ13C of Mandarina chichijimana is clearly associated with the relative distribution and abundance of C3 and C4 plants. Because C4 plants are more common near the beach than inland forest, shell δ13C was higher in the former than in the latter. A significant association between vegetation and shell δ13C was detected even in the samples from a discrete area such as the Minamizaki limestone outcrop. We also found that the carbon isotope composition of body tissue and shell carbonate are strongly correlated, confirming that variation in shell δ13C reflects variation in diet. As a wide range of carbon isotope values were found across localities and within the same sampling site, this suggests that individual snails have a varied diet.
In addition to the diet components, some percentage of foreign carbonates ingested from the surrounding sediments may contribute to the carbon isotope value (Goodfriend and Hoods, 1983; Goodfriend, 1987). Previous studies suggested that CaCO3 ingested from the soil or external water affects carbon isotope ratio in the shell of some land snail species (Goodfriend and Ellis, 2002; Ortiz et al., 2006; Yanes et al., 2008). However, an influence of ingested CaCO3 on the shell δ13C can probably be ignored in Mandarina, because no difference in shell δ13C was found between the samples from the limestone area and those from the volcanic rock area. When δ13C of the shell carbonate is derived exclusively from the consumed vegetation, the shell δ13C value is theoretically expected to correlate with the δ13C of the soft parts with a slope of 0.94847 (Stott, 2002; Yanes et al., 2008). The slope of the regression line obtained in Mandarina (0.912) is very close to this value, and is not significantly different from the expected value of the slope (P > 0.05). This fact also suggests that influence of the foreign carbonates from the soil can be neglected in Mandarina. In addition, this finding is also supported by previous studies that showed an absence of Δ14C anomalies in live M. chichijimana shells collected from the limestone areas (Chiba et al., 2007). The implication is that the frequency of C4 plants in the habitat can be estimated by shell δ13C in Mandarina.
Table 3.
Values of δ13C of the plants commonly found at the study sites. The locality where each plant specimen was collected is also shown.

Table 4.
Results of GLMM analysis on the association of shell δ13C values with frequency of C4 plants and geology (limestone or volcanic rock) of the sampling sites.

Figure 3.
Scatter plots of the isotopic composition for shells and soft parts of live specimens of Mandarina chichijimana. A regression line of the plots is also plotted.

In the present study, we pooled the plant species as either C3 or C4 plant type. Difference in δ13C value among the C3 species may affect the shell δ13C variations, because the C3 species Scaevola aemula showed particularly high δ13C value (-25.5‰). This species is restricted to the localities near the beach, and this high δ13C value may reflect the arid environment of the coastal area, because the 13C abundance of plant tissues is directly correlated with water use efficiency (WUE), hydrologic stress (e.g., Farquhar et al., 1982, 1989) or temperature (e.g., Deleens et al., 1985). Although the exact cause of the high δ13C value of this plant is unclear, the frequency of this species was associated with that of C4 plants, and therefore, the influence of heterogeneity in δ13C value among the C3 species on the spatial patterns of shell δ13C variations is negligibly small in the present case.
Higher shell δ13C in the modern samples (average δ13C, -10.2 for modern shells in Minamizaki) than in the fossil samples (average δ13C, -11.5) suggests that C3 plants were more common ∼20000 yr BP than at present in Minamizaki. This vegetational change is most likely to be a result of sea level change since the end of the last ice age. The sea level stood at least 80 m lower than the present level in Ogasawara during the last glacial maximum (LGM) (Toya et al., 1978). This implies that Minamizaki was more than 1 km inland from the coastline. Thus, lower δ13C in the fossil shells than in the modern shells from Minamizaki suggests that the habitat of this area at the LGM was inland forest, where C4 plants were absent or very rare.
An alternative explanation that can not be ruled out is that differences in the vegetation between present and LGM may reflect climatic differences. If the climate of Ogasawara in the LGM was more moist than at present, the vegetation of this period might include a lower frequency of C4 plants. However, this seems unlikely since other climate models suggest that decreased atmospheric CO2 initiated arid conditions and decreased forest cover during the period (Asahara, 1999; Shen et al., 2005). In either case, the results imply that variation in the δ13C of Mandarina shell may provide useful information to estimate habitat in the past.
Although the present study suggests that relative distribution of C3 and C4 plants is the major factor that causes variation in the shell δ13C, other factors should also affect variation in the shell δ13C. For example, a high level of geographical variation in the shell δ13C was found among the samples M3–M8 (Figure 2) that were collected from the inland areas without C4 plants. The δ13C of a leaf can be affected by soil moisture content or rainfall input (Medina and Minchin, 1980; Sparks and Ehleringer, 1997; Hanba et al., 2000; Bowling et al., 2002). In addition, isotope fractionation during decomposition and humification of organic matter may cause variations in litter δ13C (Nadelhoffer and Fry, 1988), and difference in the level of decomposition may cause differences in the average δ13C in the litter and shells among the localities. The δ13C in the CO2 of the forest also affects shell δ13C through the structure of the forest or landform. Dense forest and forest in the valley may accumulate CO2 derived from respiration of plants and soil at a greater rate than open forest or forest on the plateaus. Shell δ13C should therefore be correspondingly lower in the dense forest or forest in the valley than in the sparse forest or forest on the plateaus (Buchmann et al., 2002; de Araújo et al., 2007).
It was not possible to further investigate these issues, so details of the precise processes determining shell δ13C variation are still unclear. The present study nonetheless reveals that the Mandarina shell δ13C can generally be used as an indication of vegetative diet, and so fossil shells may be used to provide information on the past vegetation. It is hoped that the findings will serve as a basis for future studies to estimate the paleoenvironment of the Ogasawara Islands by examining fossil records of Mandarina.
Conclusion
The results of this study suggest that relative distribution of C3 and C4 plants is the major factor that causes variation in the shell δ13C of the land snails Mandarina chichijimana on Chichijima. Thus, variation in shell δ13C may provide useful information to estimate the vegetation of the Ogasawara Islands in the past. Further investigations of the relationship between shell δ13C and a number of other environmental factors are still needed. The present findings, nevertheless, reveal that the δ13C data of these fossil land snail shells is potentially very useful to estimate environments of Ogasawara during the last glacial periods.
Acknowledgements
We express our sincere thanks to I. Hayami, K. Tanabe, the South Kanto branch of the Ministry of the Environment, the Ogasawara branch office of the Tokyo metropolitan office and the Institute of Boninology for assistance in this survey, and K. Tomiyama for providing materials. This study was conducted under permit from the Agency for Cultural Affairs, Kanto Regional Forest Office and the Ministry of the Environment. This research was supported by the global environment research fund (F-051).