Introduction

The relative proportions of the stable isotopes of nitrogen (14N and 15N, expressed as δ15N) in ecosystem components can serve as a proxy of N dynamics and as a nondestructive indicator of how plants respond to environmental changes in terrestrial ecosystems (Amundson et al. 2003; Díaz et al. 2016; Robinson 2001). The two stable isotopes of N are discriminated in several fundamental biogeochemical processes that in turn are sensitive to environmental conditions (Amundson et al. 2003; Robinson 2001), so δ15N values have been widely used to reflect how environmental changes alter the ecosystem N-cycles over large scales (Craine et al. 2009; Ogaya and Peñuelas 2008; Peri et al. 2012; Swap et al. 2004).

The values of δ15N in individual plants are determined by the isotopic ratio of the external source and the redistribution of N within the plant (Evans 2001; Kolb and Evans 2002), while plant community-level δ15N values are also controlled by the relative abundance of plant species (Craine et al. 2015; Peri et al. 2012). Previous studies have shown different δ15N values between C3 and C4 photosynthetic pathways (Brown 1978; Sage and Pearcy 1987a, b). For instance, higher δ15N values in C4 plants than their C3 neighbors were found in western Australia (Wooller et al. 2005) and the Mediterranean region (Hartman and Danin 2010), but C3 plants had higher δ15N values than C4 plants in southern Africa (Aranibar et al. 2008).

Relationships between plant δ15N values and precipitation are a product of water availability and soil N sources during plant growth (Handley et al. 1999). The general pattern that soil and plant δ15N values decrease with increased precipitation has been demonstrated at both regional and global scales and suggests different biogeochemical processes and cycles of N induced by increased aridity, producing more open cycles in drier regions (Amundson et al. 2003; Aranibar et al. 2004; Craine et al. 2009; Ogaya and Peñuelas 2008). In less stressed environments the higher plant uptake allows for a greater N retention in the plant-soil system and reduced loss of 15N in a more closed N cycle (Amundson et al. 2003; Aranibar et al. 2004). However, the responses of plant δ15N to environmental changes are also dependent on the photosynthetic pathways (Murphy and Bowman 2009). The response of plant δ15N values to increasing water availability was more positive in C3 than in C4 grasses in Australian grasslands (Murphy and Bowman 2009). It has been argued that plant δ15N values are inversely correlated with precipitation for C3 but not for C4 plants in southern Africa (Swap et al. 2004).

The differential responses of δ15N values between the coexisting C3 and C4 plants to climate changes are caused by their different N sources (Wang et al. 2016; Michelsen et al. 1998; Pardo et al. 2006). It has been well known that the N sources from mycorrhizal fungi and direct root uptake from soils can vary isotopically as a result of local environmental conditions (Hobbie et al. 2000). The enhanced dependence of C3 or C4 plants on mycorrhizal fungi generally reduces their corresponding δ15N values by delivering 15N-depleted N to host plants (Hobbie and Colpaert 2003; Hobbie and Högberg 2012), and C3 or C4 plants that prefer nitrate are predicted to have lower δ15N values than those plants that prefer ammonium (Houlton et al. 2007). Moreover, changes in soil water availability may also alter δ15N values in C3 and C4 plants by changing their rooting depth and N availability with soil depth and thereby the 15N signature of plant N sources (Kahmen et al. 2008), because nitrate and ammonium sources at different soil depths can vary in δ15N signature (Hobbie and Ouimette 2009; Xue et al. 2009). Until recently, however, a lack of available ecological data has limited our ability to determine the underlying mechanisms for the differential responses of C3 and C4 photosynthetic pathways to climatic variables (Hartman and Danin 2010).

To address this knowledge gap, we investigated plant δ15N of all the species in 26 plant communities across a 3200 km climatic gradient in arid and semiarid grasslands of northern China, a suitable study region because the impact of climatic factors on ecosystem N cycles is particularly strong as water stress and N availability are the main constraints limiting plant growth and microbial activity in these areas (Bai et al. 2004; Cai et al. 2017). The C3 and C4 species are widely distributed and coexist across this transect. The dominant plant growth forms gradually changed from grasses and forbs to low shrubs with increasing aridity from the east to the west (Hilbig 1995; Ni 2003; Pyankov et al. 2000). The unique features of this region encompass relatively gentle geographical relief, distinct patterns of precipitation and temperature, and relatively low N deposition levels. We hypothesized that I) plant δ15N values would increase towards the dry end of the climatic gradient for both C3 and C4 plants and the whole plant community, and II) the response sensitivity would differ between the coexisting C3 and C4 plants given the differences in N metabolism and the large fractionations within plants (Robinson 2001). Specifically, we expected that C3 plants would be more sensitive to aridity than C4 plants, consistent with previous results in southern Africa (Swap et al. 2004) and Australian grasslands (Murphy and Bowman 2009).

Material and methods

Study sites

In early August 2012, our study was conducted along an east-west transect across arid and semiarid grasslands in northern China, which has been previously described (Wang et al. 2014; Luo et al. 2016). This transect is approximately 3200 km long and covers approximately 10° latitude and 33° longitude (39.8–50.5°N and 87.7–120.5°E) (Fig. 1). The topography of the study area consists of gently rolling hills and tablelands, with elevations ranging from 700 m in the east to 1500 m above sea level in the west. The arid and semiarid grasslands are far from human perturbations, subjected to minimal animal grazing and other anthropogenic disturbances. This region has a dry, continental climate with marked annual variation in both temperature and precipitation. Mean annual precipitation (MAP) ranges from 450 mm (east) to 50 (west) mm, and mean annual temperature (MAT) ranges from −1.5 °C (east) to 9.5 °C (west). The interaction of increasing MAP and decreasing MAT is closely tracked by species richness and vegetation cover (%), which both increase with increased water availability from the west to the east across this transect. A total of 26 sites at intervals of ca. 150 km were selected along the transect. The latitude, longitude and elevation of each sampling site were recorded by GPS (eTrex Venture, Garmin, USA).

Fig. 1
figure 1

A 3200-km long transect in arid and semiarid grasslands of northern China. A total of 26 sampling sites from west to east were selected along the aridity gradient. Two 50 m × 50 m plots were selected at each site, and five1 m × 1 m sampling subplots (or 5 m × 5 m sampling subplots in site dominated by low shrubs) were placed within each plot. Three typical vegetation types are distributed with increasing aridity: meadow steppe, typical steppe, and desert steppe. The dominant plant growth forms change gradually from perennial grasses (Leymus chinensis (C3), Stipa grandis (C3) and Cleistogenes squarrosa (C4)) to low shrubs (Calligonum mongolicum (C3) and Suaeda microphylla (C4))

Four representative types of vegetation can be found along the transect: desert steppe, typical steppe and meadow steppe, which are characterized by increasing precipitation and decreasing temperature from the west to the east (Fig. 1). The desert steppe, at the dry end of the gradient, is dominated by low shrubs Calligonum mongolicum (C3) and Suaeda microphylla (C4), with low species richness and soil organic matter content. The typical steppe, in the central part of the gradient, is dominated by Salsola collina (C4) and Reaumuria soongarica (C3). The meadow steppe, at the wet end of the gradient, is dominated by Leymus chinensis (C3), Stipa grandis (C3) and Cleistogenes squarrosa (C4) and has relatively high species richness and soil organic matter content. Related soil types of this region are gray-brown desert soils, brown calcic soils and chestnut soils distributed from west to east, belonging to the Kastanozems in the classification system of the Food and Agriculture Organization and Mollisol order of the US Soil Taxonomy.

Sampling and measurement

At each site, two 50 m × 50 m main plots were established and five 1 m × 1 m sampling subplots (or 5 m × 5 m sampling subplots in site dominated by low shrub) were placed within each main plot at the four corners and the center (Fig. 1). Plant species presence were measured in each subplots, and from these data species richness (number of plant species per subplot) were calculated. Standing crop was estimated from the dry biomass of the aboveground living parts. Aboveground biomass was sampled by clipping all plants at ground level within each sampling subplot. All living plants were sorted to species and then stored in paper bags. Plant materials were dried at 105 °C for 30 min in a portable drying oven to minimize respiration and decomposition and were later completely oven dried at 65 °C to constant weight in the laboratory. After removal of surface litter, one composite soil sample (0–10 cm depth) was randomly collected from each sampling subplot using a soil corer (2.5 cm diameter). Soil samples was carefully removed from the plant material and then separated into two sub-samples: one was stored in a cloth bag at room temperature (air-dried soils); the other one was stored in a plastic bag in a refrigerator at 4 °C (fresh soils). A detailed description of the vegetation and soil survey was documented in Luo et al. (2015, 2016).

Dried plant and soil materials were ground in a ball mill (NM200, Retsch, Haan, Germany) and stored in a plastic bag until further analysis. Plant and soil bulk δ15N values and soil total N concentrations were determined using an elemental analyzer (Elementar Vario Micro Cube, Elementar, Germany) connected to an isotope ratio mass spectrometer (IsoPrime100, Isoprime Ltd., UK), with an overall precision better than 0.2‰. δ15N values are expressed in per mil (‰) unit, relative to the atmospheric N2 standards.

Methods for the determination of soil pH and microbial-biomass N (MBN) has been described previously (Luo et al. 2016). Briefly, soil pH was measured using a pH electrode (S210 SevenCompact™, Mettler, Germany) in a 1: 2.5 mixture of soil: water. The concentration of MBN was measured with the fumigation-extraction method.

The MAT and MAP data (data range 1950–2000) were extracted from a global climate dataset with a resolution of 0.0083° × 0.0083° (approximately 1 km2 at the equator), obtained from http://www.worldclim.org. The potential evapotranspiration (PET) data (data range 1950–2000) were extracted from the CGIAR-CSI Global Aridity Index and Global Potential Evapo-Transpiration Climate Database (http://www.cgiar-csi.org/data/global-aridity-and-pet-database). Aridity (unitless) was calculated as 1-AI, where AI, the ratio of MAP to PET, is the aridity index (Luo et al. 2016). Aridity therefore increased with increasing MAT and decreasing MAP. Across this transect, aridity ranged from 0.45 to 0.95, equivalent to a range in MAP of 450–50 mm, and a range in MAT ranged from −1.5 to 9.5 °C. The aridity was applied to incorporate MAP and MAT into one parameter to assess the variations in plant δ15N values along the climatic gradient due to the strongly positive correlation between PET and MAT.

Data analysis

All sampled plant species were classified into C3 or C4 photosynthetic pathways. If the specimen could be assigned to a genus, classification was made using the identification in Watson and Dallwitz (1992). If the specimen could not be identified to generic level, classification was made by the δ13C values (Cerling et al. 1997).

Plant community δ15N values were defined as the overall mean of δ15N values across all species (n, species richness) weighted by the relative (fractional) contribution of each species to the overall biomass at each quadrat (Kichenin et al. 2013):

$$ \mathrm{Community}\ {\updelta}^{15}\mathrm{N}=\left({\mathrm{biomass}}_1\times {\updelta}^{15}{\mathrm{N}}_1+{\mathrm{biomass}}_2\times {\updelta}^{15}{\mathrm{N}}_2+\dots \dots {\mathrm{biomass}}_{\mathrm{n}}\times {\updelta}^{15}{\mathrm{N}}_{\mathrm{n}}\right)/\mathrm{total}\ \mathrm{biomass}. $$

Ordinary least squares (OLS) regression was used to analyze the responses of plant community δ15N values to increasing aridity. Binary regression was conducted to analyze the relationship between aridity and soil δ15N values. Then, OLS regression was also used to examine the correlation between mean community and soil δ15N values. To further analyze patterns of plant δ15N values, OLS regression was applied to test the relationships of plant δ15N values and N concentrations with aridity for both C3 and C4 plants. Then, OLS regression was applied to test the relationships of plant δ15N values with species richness for both C3 and C4 plants. We conducted analysis of covariance to identify the differences in the slopes of the regression lines between C3 and C4 photosynthetic pathways. Steeper slope means greater sensitivity of δ15N value in this plant type in response to aridity.

To examine the underlying mechanism under the different responses of δ15N values to aridity between coexisting C3 and C4 plants, structural equation modeling (SEM) was applied to examine the interactive effects of climatic and soil variables on the δ15N values in C3 and C4 plants, respectively. In the SEM analysis, we compared the model-implied variance-covariance matrix against the observed variance-covariance matrix. Data were fitted to the models using the Akaike information criterion and the goodness of fit index. For simplicity, the least significant path was deleted and the model was re-estimated; then the next least significant path was removed, and so on, until the paths that remained in the final SEM were all significant. Standard errors and the significance level (P value) were calculated using bootstrapping (1200 repetitions).

All statistical analyses were carried out using the statistical package of SPSS 13.0 for Windows® (SPSS Inc., Chicago, IL, USA, 2004) and the sem function in the sem package of R-project (R i386 3.1.1).

Results

Plant δ15N values significantly increased with increasing aridity at the community level (P < 0.001, Fig. 2). Removing the leguminous species from this analysis did not change the results (data not shown). The relative biomass of leguminous species (%) showed no significant relationship with aridity (data not shown). Plant communities in drier sites contained a greater percentage of total plant biomass of species with higher δ15N values (Table 1). Soil δ15N values increased first and then reduced with increasing aridity (Figure S1) and was nonlinearly associated with plant community δ15N values (Fig. 3). Removing three sites with soil δ15N values >10 ‰ (outliers) did not change the results (see the insets in Figs. 3 and S1).

Fig. 2
figure 2

Correlation between aridity and plant community N isotopic signature (δ15N) along the grassland transect in northern China. Plant community δ15N values were defined as the overall mean of δ15N values across all species weighted by the relative contribution of each species to the overall biomass. Aridity was calculated as 1-AI, where AI, the ratio of precipitation to potential evapotranspiration, is the aridity index

Table 1 The most abundant species of each community type
Fig. 3
figure 3

Relationships between plant community and soil bulk δ15N along the grassland transect in northern China. Plant community δ15N values were defined as the overall mean of δ15N values across all species weighted by the relative contribution of each species to the overall biomass. When the three sites with soil δ15N values higher than 10 ‰ were removed, the non-linear relationship remained between plant and soil δ15N values

Plant δ15N values also significantly increased with aridity in both C3 and C4 plants at the level of individual species (P < 0.001, Fig. 4). The interaction between aridity and type of photosynthetic pathway was significant, i.e., the slope of the regression line for aridity and δ15N values was significantly steeper for C4 than C3 plants (P < 0.001). When the specific-biomass weighting factor was considered, the response of δ15N values to increasing aridity was also more positive in the C4 than in the C3 plant group (Figure S2). Plant N content did not show any significant relationships with aridity for both C3 and C4 plants along the transect (Figure S3). Plant δ15N values reduced with increasing species richness in both C3 and C4 plants, and the slope of the regression line for species richness and δ15N values was significantly steeper for C4 than C3 plants (P < 0.001, Figure S4). The proportional contribution of C3 plants to total biomass reduced and that of C4 plants increased with the increase in aridity (Figure S5). Soil pH increased with increasing aridity, while standing crop, species richness, soil total N concentration, soil C:N ratio, and MBN concentration all reduced with increasing aridity along the transect (Figure S6).

Fig. 4
figure 4

Correlation between aridity and N isotopic signature (δ15N) for C3 and C4 plants along the grassland transect in northern China. Aridity was calculated as 1-AI, where AI, the ratio of precipitation to potential evapotranspiration, is the aridity index

The SEM analyses showed that the total effects of aridity, soil pH and ANPP were positive on δ15N values in both C3 (Fig. 5a) and C4 plants (Fig. 5b). Aridity indirectly affected plant δ15N values via the positive effect on soil pH and the negative effects on ANPP and species richness for both C3 (Fig. 5a) and C4 plants (Fig. 5b). The total effects of soil total N and MBN concentrations were negatively and positively correlated with δ15N values in C4 plants, respectively (Fig. 5b). Aridity indirectly affected plant δ15N values via the negative effects on soil total N and MBN concentrations for C4 plants (Fig. 5b).

Fig. 5
figure 5

Diagram of the structural equation modeling (SEM) that best explain the maximum variance of (a) C3 and (b) C4 plant N isotopic signatures (δ15N) along the environmental gradient in northern China. Numbers adjacent to arrows are standardized path coefficients, analogous to relative regression weights, and indicative of the effect size of the relationship. Dashed and continuous arrows indicate negative and positive relationships, respectively. Arrow width is proportional to the strength of the relationship. Goodness-of-fit statistic for each model are shown in the lower right corner. The proportion of variance explained (R2) appears alongside every response variable in the model. * P < 0.05, ** P < 0.01, *** P < 0.001. Aridity was calculated as 1-AI, where AI, the ratio of precipitation to potential evapotranspiration, is the aridity

Discussion

N stable isotopes in plants along the aridity gradient

Consistent with our hypothesis, plant δ15N values of both the C3 and C4 functional groups and the entire community increased towards the dry end of the climatic gradient across northern China’s grasslands, where N and water availability are two of the most constraining factors limiting plant growth and microbial activity. This finding is similar to that reported in other continents (Aranibar et al. 2008; Austin and Sala 1999; Heaton 1987; Swap et al. 2004) and in an extensive global synthesis (McCulley et al. 2009; Murphy and Bowman 2009; Schulze et al. 1998) conducted at the level of species rather than the entire plant community. Plant δ15N values also increased with aridity in our study in the entire plant community, similar to individual species, which was due to the substitution of plant species/functional groups with lower δ15N by those with higher δ15N when aridity rose along the climatic gradient (see Table 1). Higher plant δ15N values indicate a lower capacity of plants to retain N, because the lighter 14N isotope is more easily cycled and lost (Dalal et al. 2013; McCulley et al. 2009). Plant species that are less efficient in retaining N are favored when aridity increases along the climatic gradient. This result suggests that water conservative mechanisms constitute a trade off with the capacity to retain and use N. This trade-off has been observed and commented in previous studies (Dijkstra et al. 2016). Thus, the biological response to drier conditions when water limitation is the main driver of adaptive responses provokes a decrease in the capacity to retain N in the plant community.

The increased plant δ15N values with aridity, however, do not agree with our previous study along the same gradient, which found that the ecosystem N-cycles, based on soil bulk δ15N rather than plant δ15N signals, were more closed at the two ends of the aridity gradient and more open in the middle of the aridity gradient (Wang et al. 2014). Similarly, Díaz et al. (2016) recently reported that soil δ15N values increased with increasing aridity along an elevational/climatic gradient in northern Chile, as expected, whereas plant δ15N values had a hump-shaped relationship with increasing aridity. The decoupled relationship between the δ15N values of plants and soils indicates the differences in biogeochemical processes underlying N dynamics between vegetation and soil. An increase in plant δ15N values with aridity, independent of soil bulk δ15N values, may be caused by changes in the uptake of nitrate versus ammonium along the aridity gradient (Houlton et al. 2007; Takebayashi et al. 2010). Species that prefer nitrate relative to ammonium generally have lower δ15N values than species that prefer ammonium (Houlton et al. 2007). In our previous study along the same climatic gradient, we found that δ15N values of ammonium consistently increased, while δ15N values of nitrate increased first and then decreased with increasing aridity (see Liu et al. 2017). These results suggest a shift of dominant inorganic N sources for plants with aridity, resulting in a positive relationship between aridity and plant δ15N values and a decoupled relationship between plant and soil bulk δ15N values. Similarly, Houlton et al. (2007) demonstrated that increased aridity resulted in a switch in the dominant N source for plant growth, leading to a reduction in plant δ15N values from drier to wetter sites in tropical forest ecosystems.

N isotope and N use in C4 versus C3 plants

In line with our hypothesis, the responses of plant δ15N values to environmental changes differed between the coexisting C3 and C4 photosynthetic pathways (Fig. 5). The δ15N values in C4 plants were positively correlated with soil MBN concentration and negatively correlated with soil total N concentration but not in C3 plants (Fig. 5). This suggests a stronger competition intensity of N uptake between soil microbes and C4 plants when soil N content becomes scarcer under drier conditions (Liu et al. 2016; Ouyang et al. 2016). The increases in plant δ15N values in C4 plants along the aridity gradient therefore corresponded to a large proportion of soil N incorporated into microbial biomass, suggesting that C4 plants competed with soil microbes for N less strongly than C3 plants (Liu et al. 2016; Ouyang et al. 2016)). Moreover, our results showed that the reduction in species richness with increasing aridity directly resulted in a reduction in plant δ15N values in both C3 and C4 plants, with the effects being stronger in C4 than C3 plants (Figure S4). These results suggest that C4 plants were more sensitive to plant neighborhood competition with respect to N uptake than C3 plants (Harrison et al. 2007; Mariotte et al. 2013). Taken together, C4 plants appeared more affected by competition pressure of neighboring plants and soil microbes than C3 plants with respect to N uptake and C3 plants had an apparent advantage in the use and retention of N compared to C4 plants in arid and semiarid grasslands.

This advantage of C3 plants in the use of N may improve their competitive ability and thus their survival capacity under dry conditions. Previous studies have reported that C3 plants would be replaced by C4 plants due to their lower water-use efficiency if the global climate becomes drier in the near future (Pyankov et al. 2000; Wittmer et al. 2010). However, this higher N-use efficiency we observed in C3 plants could counteract the competitive advantage of the more water-use efficient C4 plants under drier conditions, thereby partially avoiding the total replacement of C3 by C4 plants. A reduction in transpiration can further decrease the capacity of C4 plants to take up and compete for nutrients (Cramer et al. 2009). Our results provide new evidence of a trade-off between N-use and water-use efficiencies for plants with different photosynthetic pathways (Dijkstra et al. 2016). The higher ability of C3 plants to absorb, retain and use N could, at least partially, explain why C3 plants are not fully replaced by the more water-use efficient C4 plants under arid conditions when N supply decreases.

Conclusion

Our results have two important implications for predicting the responses of vegetation and biogeochemical cycles to climate change. First, plant δ15N values of both the C3 and C4 functional groups and the entire community increased towards drier conditions due to an increase in δ15N of each individual species, and also to an increased dominance of species with higher δ15N values. Along the aridity gradient, plant δ15N values did not covary with soil bulk δ15N values, resulting from a shift in dominant inorganic N sources for plant growth along the aridity gradient. The decoupled relationships suggest that the mechanisms underlying soil bulk δ15N patterns should be carefully applied to plant δ15N patterns in arid and semiarid ecosystems. Second, our results suggest that competition pressure for N by neighboring plants and soil microbes became more intense for C4 than C3 plants, thereby partly counteracting the competitive advantage of C4 plants due to their higher water use efficiencies under drier and warmer conditions. These findings provide new hypotheses to explain why C3 plants are not completely replaced be C4 plants in drier and warmer conditions.