15N Natural Abundance of C3 and C4 Herbaceous Plants and Its Response to Climatic Factors along an Agro-Pastoral Zone of Northern China

The nitrogen isotope composition of plants (δ15N) can comprehensively reflect information on climate change and ecosystems’ nitrogen cycle. By collecting common herbs and soil samples along the 400 mm isoline of mean annual precipitation (MAP) in the agro-pastoral zone of North China (APZNC) and measuring their δ15N values, the statistical characteristics of foliar δ15N of herbs and the responses of foliar δ15N to the MAP and mean annual temperature (MAT) were analyzed. The results showed that: (1) the δ15N values of all herbs investigated varied from −5.5% to 15.25%. Among them, the δ15N value range of C3 herbs (−5.5~15.00%) was wider than that of C4 herbs (−2.17~15.25%), but the average value (3.27%) of C3 herbs was significantly lower than that of C4 herbaceous plants (5.55%). This difference provides an important method for identifying plants of different photosynthetic types by nitrogen isotope technology. (2) Along the transect from northeast to southwest, the δ15N of both C3 and C4 herbs decreased with the increase in the MAP, but not significantly for C3 herbs. The inverse relationship between the nitrogen isotopic signatures of herbs and MAP is consistent with previous studies. However, the MAP in the APZNC is found to only explain a small amount of the observed variance in the δ15N herbs (C3 herbs: 10.40%; C4 herbs: 25.03%). (3) A strong negative relationship was found between δ15N of herbs and MAT across the transect (C3 herbs: −0.368%/°C; C4 herbs: −0.381%/°C), which was contrary to the global pattern and some regional patterns. There was no significant difference in the δ15N responses of two different photosynthetic herbs to temperature, but the effect of temperature on the variances of δ15N of C3 and C4 herbs was significantly greater than that of precipitation. This suggests that temperature is a key factor affecting foliar δ15N of herbs in this transect. The above findings may be of value to global change researchers studying the processes of the nitrogen cycle and gaining an insight into climate dynamics of the past.


Introduction
Nitrogen (N) has long been considered as one of the most extensive nutrients that constrain plant growth, maintenance, and reproduction in many terrestrial ecosystems, and its cycle affects almost all aspects of ecosystem functions [1][2][3]. Since the natural abundance of 15 N in plant tissue is a combined result of a series of biogeochemical processes and environmental changes, the nitrogen isotope composition (δ 15 N) in plants can record a series of climate and environmental information related to physio-ecological processes of plants to a certain extent, and become an focal tool to infer past short-term (e.g., annual time scales) variations in the ecological environment or to indirectly indicate ecosystem function and N-cycling processes that are difficult to measure directly [2,[4][5][6][7][8][9]. These applications are based mainly on the general response pattern of plant δ 15 N to ecological and climatic This can further promote our understanding of how foliar δ 15 N in herb species responds to environmental factors and help us explain the changes in functions and regulatory mechanisms of key factors involved in the N cycling of grassland ecosystems. Contrast to previous studies, our research has made important contributions in three aspects. First, by comparing the differences in δ 15 N values between C3 and C4 herbs, a new method is provided for identifying plants with different photosynthetic types by nitrogen isotope technology. Second, based on the response of nitrogen isotopes of C3 and C4 herbs to temperature and precipitation changes (if this pattern is indeed robust), it can provide basic data for using stable isotopes to explain the process of the nitrogen cycle in temperate grasslands of North China. Third, because the δ 15 N of C3 and C4 herbs contains a lot of climate-environmental information, this will provide a new idea for paleoclimatologists to use plant δ 15 N as a climate proxy to deeply explore the past climate dynamics and paleoenvironment reconstruction.
The agro-pastoral zone of Northern China (FPZNC) is a relatively independent geographical region. The complex and diverse topography of the study area consists of gently rolling hills, grasslands, sandy land, and platforms, with an altitude of 650 m to 1800 m. Its natural environment variables such as climate, vegetation and soil have distinct transitional characteristics. The sensitivity of the geochemical cycle within the ecosystem to temperature and humidity fluctuations makes this region an obvious indicator of environmental changes. This region belongs to a typical temperate continental semi-arid monsoon climate: dry and hot in summer, cold in winter, sunny in autumn, strong winds in spring and frequent sandstorms. The mean annual precipitation (MAP) ranges between 345 mm and 443 mm and the mean annual temperature (MAT) varies from −6.1 • C to 8.9 • C, with a decreasing trend from northeast to southwest [29]. In this area, more than 60% of the annual rainfall occurs in the summer season from June to August owing to its climate condition and geographical location. The natural vegetation is consistent with semi-humid, semi-arid and semi-desert climates extending from the northeast to the southwest, and is expressed as continuous changes in vegetation from meadow steppes to typical steppes and desert steppes from east to west. The C3 and C4 herbs with different photosynthetic pathways coexist in the FPENC, of which C3 herbs are dominant and widely distributed, and C4 herbs are limited in number [29]. The soil types formed under unique climatic conditions are mainly chestnut soil, loess soil and chernozem. The unique features of this region, including a continuum of mesic to xeric grassland types, distinct climatic gradients and relatively light human disturbance, provide ideal conditions to explore the response patterns of the natural abundance of 15 N in C3 and C4 herbs along a regional environmental gradient. Although some scholars have explored the relationship between plant δ 15 N and environmental factors in the APZNC, the scope of the research is limited to a typical section [30,31]. In such an agro-pastoral ecotone that is thousands of kilometers long, there are few reports on the systematic investigation of the relationships between δ 15 N of C3 and C4 herbs and environmental variables. This study systematically investigates the variations in δ 15 N of C3 and C4 herbs along a precipitation and temperature gradient in the APZNC, and addresses two scientific questions: (1) whether are there significant differences between the δ 15 N of C3 and C4 herbs from the same sites in the APZNC due to differences between photosynthetic pathways? (2) How do the δ 15 N values of C3 and C4 herbs in this area respond to driving factors such as changes in temperature and precipitation, and is there any difference in the response patterns?

Comparison of Foliar δ 15 N between C3 and C4 Herbs
As shown in Figure 1, the frequency distribution of foliar δ 15 N values of C3 and C4 herbs in the APZNC was of unimodal type. The foliar δ 15 N values of overall herbs ranged from −5.50% to 15.25%, with a coefficient of variation (CV) of 1.05, indicating that the δ 15 N values of herbs had a large spatial variability in the study area. Among them, the C3 herbs showed a wider range of δ 15  the CV of the former was 1.9 times that of the latter, which might be related to the fact that the number of C3 herb species collected in this study was far more than that of C4 herb species (Table 3). The reason is that in the APZNC, although C3 and C4 herbs are widely distributed, the number of C4 herbs is extremely limited. Further analysis found that in our nitrogen isotope data set, more than 95.0% of the foliar δ 15 N values fell in the range of −4.0% to 12.0%, which is basically consistent with the previously reported range of plant δ 15 N values in North China (C3 plants: −5.1-13.0%; C4 plants: −3.2-12.4%) [11]. However, compared with the range of foliar δ 15 N for 11,000 plants worldwide from −10.0% to 17.0% [10], the range of foliar δ 15 N investigated in the APZNC is much more concentrated. There may be two reasons for this phenomenon. First, the samples in this study were all from the transition zone between the semi-arid region and the semi-humid region, so the climate conditions in our study area were relatively simple compared with those in the rest of the world. Second, the plant samples collected in this study were all herb species, while the ones from all over the world included trees, shrubs and herbs. Generally, plants with different life forms have selectivity for the absorption of different nitrogen sources in soil, which leads to significant differences in δ 15 N among plants with different life forms, manifested as arbor > shrub > herb [4,32].

Comparison of Foliar δ 15 N between C3 and C4 Herbs
As shown in Figure 1, the frequency distribution of foliar δ 15 N values of C3 and C4 herbs in the APZNC was of unimodal type. The foliar δ 15 N values of overall herbs ranged from −5.50% to 15.25%, with a coefficient of variation (CV) of 1.05, indicating that the δ 15 N values of herbs had a large spatial variability in the study area. Among them, the C3 herbs showed a wider range of δ 15 N values (−5.50-15.00%) than C4 herbs (−2.17-15.25%), and the CV of the former was 1.9 times that of the latter, which might be related to the fact that the number of C3 herb species collected in this study was far more than that of C4 herb species (Table 3). The reason is that in the APZNC, although C3 and C4 herbs are widely distributed, the number of C4 herbs is extremely limited. Further analysis found that in our nitrogen isotope data set, more than 95.0% of the foliar δ 15 N values fell in the range of −4.0% to 12.0%, which is basically consistent with the previously reported range of plant δ 15 N values in North China (C3 plants: −5.1-13.0%; C4 plants: −3.2-12.4%) [11]. However, compared with the range of foliar δ 15 N for 11,000 plants worldwide from −10.0% to 17.0% [10], the range of foliar δ 15 N investigated in the APZNC is much more concentrated. There may be two reasons for this phenomenon. First, the samples in this study were all from the transition zone between the semi-arid region and the semi-humid region, so the climate conditions in our study area were relatively simple compared with those in the rest of the world. Second, the plant samples collected in this study were all herb species, while the ones from all over the world included trees, shrubs and herbs. Generally, plants with different life forms have selectivity for the absorption of different nitrogen sources in soil, which leads to significant differences in δ 15 N among plants with different life forms, manifested as arbor > shrub > herb [4,32]. In the present study, the average foliar δ 15 N value of all investigated herbs was 4.06‰ (n = 231). Among them, the average value of δ 15 N of C3 herbs was 3.27‰ (n = 151), which was significantly lower (p < 0.01) than the average value of 5.55% of C4 herbs (n = 80). Within the same sampling location, the δ 15 N value of C4 herbs was significantly higher than that of C3 herbs at most sampling sites (Figure 2), illustrating a different N use by the two types of herbs with different photosynthetic pathways. The higher δ 15 N values for C4 herbs than for C3 herbs were consistent with previous studies [20,24,25,28]. For instance, the results from the secondary grassland in South China and the arid and semiarid grasslands of North China showed that C3 plant δ 15 N values were significantly more depleted than C4 plant δ 15 N values [20,24]. The reasons why the δ 15 N value of C4 herbs In the present study, the average foliar δ 15 N value of all investigated herbs was 4.06‰ (n = 231). Among them, the average value of δ 15 N of C3 herbs was 3.27‰ (n = 151), which was significantly lower (p < 0.01) than the average value of 5.55% of C4 herbs (n = 80). Within the same sampling location, the δ 15 N value of C4 herbs was significantly higher than that of C3 herbs at most sampling sites (Figure 2), illustrating a different N use by the two types of herbs with different photosynthetic pathways. The higher δ 15 N values for C4 herbs than for C3 herbs were consistent with previous studies [20,24,25,28]. For instance, the results from the secondary grassland in South China and the arid and semi-arid grasslands of North China showed that C3 plant δ 15 N values were significantly more depleted than C4 plant δ 15 N values [20,24]. The reasons why the δ 15 N value of C4 herbs was higher than that of C3 herbs could be explained from the following two aspects: On one hand, foliar δ 15 N is usually affected by the difference in carbon and nitrogen metabolism between different photosynthetic plants. Due to the differences in the photosynthetic nitrogen utilization rate between C3 and C4 plants, there are also differences in N-use efficiency. In general, C3 plants have more advantages in using nitrogen than C4 plants under drier conditions [26,33]. Plants with lower N-use efficiency have higher N concentrations and higher δ 15 N values [27,34]. As shown in Figure 3, the C:N ratios of C4 herbs were significantly lower than those of C3 herbs at most sampling sites, which indicated that C4 herbs have lower N utilization efficiency and higher foliar N concentrations than C3 herbs under similar N supply conditions. This was confirmed by findings of previous studies that soil organic matter derived from C4 grasses has a faster decomposition rate (attributable to lower C:N ratios) than those from C3 grasses [35]. Hence, within the same sampling site, it is not unexpected that C4 herbs are likely to show higher δ 15 N values than C3 herbs due to the lower leaf C:N ratios of C4 herbs. Additionally, there were significant differences in leaf δ 15 N among the three C4 species. The most conspicuous difference was that the Amaranthus retroflexus had substantially higher δ 15 N (8.31 ± 2.47‰, expressed by mean ± standard deviation, the same below) than the other two species (Salsola collina: 5.31 ± 2.72‰; Setaria viridis: 4.49 ± 2.92‰). Similarly, differences in nitrogen metabolism may also contribute to the differences we found in δ 15 N between the C4 species. We found that there were clear differences in leaf nitrogen content between them (Amaranthus retroflexus: 39.27 ± 7.64%; Salsola collina: 33.93 ± 6.74%; Setaria viridis: 25.30 ± 7.55%), implying that there is an obvious difference in nitrogen metabolism among the three C4 species. On the other hand, the δ 15 N values of individual plants are also determined to some extent by the isotopic ratio of the external source. Different photosynthetic plants usually have a preference for the uptake of available nitrogen sources (e.g., ammonium nitrogen and nitrate nitrogen) from soils, and the isotopic compositions of various nitrogen sources in the soils are obviously different as a result of local environmental conditions [36]. When there is more inorganic nitrogen, especially ammonium nitrogen, held by microorganisms in the soil, the availability of nitrogen sources (e.g., ammonium nitrogen) preferentially absorbed by plants is reduced. At this time, C3 plants will change their selection from ammonium nitrogen as the main nitrogen source to nitrate nitrogen as the main nitrogen source. The isotopic analysis of Aranibar et al. indicated that the δ 15 N of nitrate in soils was lower than that of ammonium at the same sites [4]. If C3 and C4 herbs prefer nitrate and ammonium, respectively, then the δ 15 N of C3 herbs would be lower than those of C4 herbs, as it was observed in most of the sampling sites ( Figure 2). It has been reported that in some European grasslands, plants that preferred nitrate relative to ammonium had lower foliar δ 15 N than ones that preferred ammonium under controlled conditions [2]. However, the opposite result also appeared, namely plant species that preferred nitrate were more enriched in 15 N [37]. One possible explanation is that nitrate in the soil may be more enriched than ammonium owing to the loss of gaseous N after nitrification. Moreover, changes in soil water may alter δ 15 N values in C3 and C4 plants by affecting their rooting depth and N availability and, thereby, the 15 N signature of plant N sources, because nitrate and ammonium sources at different soil depths can vary in δ 15 N signature [8]. Other reports have indicated that the relative abundance of plant species in ecosystems may affect the composition and distribution of labile and recalcitrant N pools by changing the amount and quality of litter inputs, thereby changing soil nitrogen sources and affecting δ 15 N of C3 and C4 plants [35]. Unfortunately, we have not measured the ecological data in this regard, which limits our ability to determine the potential mechanism of nitrogen isotope differences between C3 and C4 photosynthetic pathways. In addition, we found that the δ 15 N of herbs decreased with the increase in soil C: N ratio along the transect, and the δ 15 N values of C3 herbs decreased more than those of C4 herbs under similar environmental conditions ( Figure 4). This means that in the APZNC, nitrogen isotopes of different photosynthetic herbs have different responses to changes in soil C: N ratio. There are reports that if the soil C:N ratio is too high, the microbial decomposition and mineralization is slow, and the available nitrogen in the soil is consumed more, thus reducing the soil available nitrogen that can be absorbed by plants, and causing low soil 15 N enrichment [38,39]. However, the response mechanism of δ 15 N for C3 and C4 herbs to the soil C: N ratio is still unclear. Since the δ 15 N values of nitrate nitrogen (NO − 3 ) and ammonium nitrogen (NH + 4 ) in the soils were not determined in this study, it limited further explanation on the causes of variation in plant δ 15 N. Therefore, it is necessary to study the effect of different nitrogen sources on plant δ 15 N in the future. and mineralization is slow, and the available nitrogen in the soil is consumed more, thus reducing the soil available nitrogen that can be absorbed by plants, and causing low soil 15 N enrichment [38,39]. However, the response mechanism of δ 15 N for C3 and C4 herbs to the soil C: N ratio is still unclear. Since the δ 15 N values of nitrate nitrogen (NO ) and ammonium nitrogen (NH ) in the soils were not determined in this study, it limited further explanation on the causes of variation in plant δ 15 N. Therefore, it is necessary to study the effect of different nitrogen sources on plant δ 15 N in the future.  Table 3. All the values are represented as mean ± SD (standard deviation) at each sampling site. Different letters at each site indicate significant differences according to Duncan's single-factor variance test at the 5% level.  Table 3. All the values are represented as mean ± SD (standard deviation) of each sampling site. Different letters at each site indicate significant differences according to Duncan's single-factor variance test at the 5% level.  Table 3. All the values are represented as mean ± SD (standard deviation) at each sampling site. Different letters at each site indicate significant differences according to Duncan's single-factor variance test at the 5% level. and mineralization is slow, and the available nitrogen in the soil is consumed more, thus reducing the soil available nitrogen that can be absorbed by plants, and causing low soil 15 N enrichment [38,39]. However, the response mechanism of δ 15 N for C3 and C4 herbs to the soil C: N ratio is still unclear. Since the δ 15 N values of nitrate nitrogen (NO ) and ammonium nitrogen (NH ) in the soils were not determined in this study, it limited further explanation on the causes of variation in plant δ 15 N. Therefore, it is necessary to study the effect of different nitrogen sources on plant δ 15 N in the future.  Table 3. All the values are represented as mean ± SD (standard deviation) at each sampling site. Different letters at each site indicate significant differences according to Duncan's single-factor variance test at the 5% level.  Table 3. All the values are represented as mean ± SD (standard deviation) of each sampling site. Different letters at each site indicate significant differences according to Duncan's single-factor variance test at the 5% level.  Table 3. All the values are represented as mean ± SD (standard deviation) of each sampling site. Different letters at each site indicate significant differences according to Duncan's single-factor variance test at the 5% level. It is worth noting that the results of this study are contrary to those obtained from South Africa, the Mediterranean region and the Tengger Desert of China, that is, C3 plants had significantly higher δ 15 N values than C4 plants in the above areas [4,23,24]. For instance, in the Tengger Desert of China, Zhao et al. reported that δ 15 N was higher in C3 plants (varying from −4.45 to 3.66‰) than in C4 plants (varying from −7.56 to 1.08) [24]. However, we think it is difficult to compare the results of the two, mainly for two reasons. Firstly, all samplings in our study were conducted along a certain temperature gradient under the condition of similar precipitation (e.g., 400 mm isoline of MAP); whereas the sampling sites selected in South Africa, the Mediterranean region and the Tengger Desert of China were set along a certain precipitation gradient [4,23,24]. Therefore, the climatic conditions of the two were quite different. Secondly, in previous studies, except for the fact that the C4 plants analyzed were herbs, plants with the C3 pathway of photosynthesis were predominantly trees and shrubs, which was obviously different from C3 herbs investigated in this study. In order to adapt to the changes in environmental conditions, plants with different life forms usually show great differences in morphology and physiological traits, which further affect the foliar δ 15 N of plants. For example, the herbs in humid, semi-arid and mesic sites had lower δ 15 N values than those of trees from the same locations [40]. Therefore, from previous studies, it is difficult to distinguish the direct effects of the photosynthetic pathway and growth form on plant δ 15 N.

Response of Foliar δ 15 N Values of Herbs to MAP
In the FPENC, there was a negative correlation between foliar δ 15 N and MAP in both C3 and C4 herbs (Figure 5a,b). That is, foliar δ 15 N values of C3 and C4 herbs showed a linearly decreasing trend with increasing precipitation (significant for C4 herbs but not significant for C3 herbs). However, the δ 15 N responses of C3 and C4 herbaceous plants to precipitation changes were significantly different within the study area. Regression analysis showed that the slope of the relationship between δ 15 N and MAP was steeper in C4 herbs compared with C3 herbs. Over the entire range of MAP, for every 100 mm increase in MAP, the δ 15 N value of C3 herbs declined by approximately 4.58% while that of C4 herbs decreased by about 6.13%, indicating that the δ 15 N values of C3 herbs were less responsive to drought than those of C4 herbs. Figure 6 showed that, on the study transect, the change rate of foliar δ 15 N of C4 herbs with AI was significantly greater than that of C3 herbs. This can also be confirmed by the results of partial correlation analysis between δ 15 N values of herbs and climate factors. Table 1 shows that the partial correlation between the δ 15 N value of C4 herbs and the MAP in our study is significantly better than that between the δ 15 N value of C3 herbs and the MAP. The reason may be that the two plant It is worth noting that the results of this study are contrary to those obtained from South Africa, the Mediterranean region and the Tengger Desert of China, that is, C3 plants had significantly higher δ 15 N values than C4 plants in the above areas [4,23,24]. For instance, in the Tengger Desert of China, Zhao et al. reported that δ 15 N was higher in C3 plants (varying from −4.45 to 3.66‰) than in C4 plants (varying from −7.56 to 1.08) [24]. However, we think it is difficult to compare the results of the two, mainly for two reasons. Firstly, all samplings in our study were conducted along a certain temperature gradient under the condition of similar precipitation (e.g., 400 mm isoline of MAP); whereas the sampling sites selected in South Africa, the Mediterranean region and the Tengger Desert of China were set along a certain precipitation gradient [4,23,24]. Therefore, the climatic conditions of the two were quite different. Secondly, in previous studies, except for the fact that the C4 plants analyzed were herbs, plants with the C3 pathway of photosynthesis were predominantly trees and shrubs, which was obviously different from C3 herbs investigated in this study. In order to adapt to the changes in environmental conditions, plants with different life forms usually show great differences in morphology and physiological traits, which further affect the foliar δ 15 N of plants. For example, the herbs in humid, semi-arid and mesic sites had lower δ 15 N values than those of trees from the same locations [40]. Therefore, from previous studies, it is difficult to distinguish the direct effects of the photosynthetic pathway and growth form on plant δ 15 N.

Response of Foliar δ 15 N Values of Herbs to MAP
In the FPENC, there was a negative correlation between foliar δ 15 N and MAP in both C3 and C4 herbs (Figure 5a,b). That is, foliar δ 15 N values of C3 and C4 herbs showed a linearly decreasing trend with increasing precipitation (significant for C4 herbs but not significant for C3 herbs). However, the δ 15 N responses of C3 and C4 herbaceous plants to precipitation changes were significantly different within the study area. Regression analysis showed that the slope of the relationship between δ 15 N and MAP was steeper in C4 herbs compared with C3 herbs. Over the entire range of MAP, for every 100 mm increase in MAP, the δ 15 N value of C3 herbs declined by approximately 4.58% while that of C4 herbs decreased by about 6.13%, indicating that the δ 15 N values of C3 herbs were less responsive to drought than those of C4 herbs. Figure 6 showed that, on the study transect, the change rate of foliar δ 15 N of C4 herbs with AI was significantly greater than that of C3 herbs. This can also be confirmed by the results of partial correlation analysis between δ 15 N values of herbs and climate factors. Table 1 shows that the partial correlation between the δ 15 N value of C4 herbs and the MAP in our study is significantly better than that between the δ 15 N value of C3 herbs and the MAP. The reason may be that the two plant types have a different N metabolism because the differences in carboxylation reactions lead to disparate isotopic fractionation between the two photosynthetic pathways (C3 and C4 plants). In addition, due to the different photosynthetic pathways' environmental controls, the N absorption by C4 plants might be more affected by the competitive pressure of neighboring plants and soil microorganisms than C3 plants along the transect, resulting in the variation in C4 plants' δ 15 N being more sensitive to precipitation [26]. This demonstrates that the responses of plant δ 15 N to environmental changes may also be dependent on the photosynthetic pathways [41]. Although the δ 15 N value of C3 and C4 herbs was correlated with the MAP, the simple relationship between δ 15 N signal and MAP only explained a small amount of the observed variance in the transect. The MAP accounted for just 10.40% and 25.03% of the nitrogen isotope variations in C3 and C4 herbs, respectively (Figure 5a,b). This means that δ 15 N signature of herbs has a limited indicative significance for the variation in precipitation across the study region. Multiple regression analysis ( Table 1) also shows that the regression coefficients of the MAP against plant δ 15 N values are only −0.035 (C3 herbs) and −0.053 (C4 herbs), respectively, which again indicates that the influence of precipitation on the δ 15 N of herbs is limited in the study transect. This phenomenon may be related to the fact that there is no obvious precipitation gradient on the study transect. This is because although precipitation is considered one of the key factors to determine plant δ 15 N in arid and semi-arid environments, most of the plant samples in this study were collected along the 400 mm isoline of MAP, and the MAP between the sampling sites varies from 345.0 to 442.8 mm, with an average value of 394.77 ± 24.07 mm. Therefore, there is no significant difference in the effect of precipitation on foliar δ 15 N of herbs due to little difference in precipitation among various sampling sites on the study transect. types have a different N metabolism because the differences in carboxylation reactions lead to disparate isotopic fractionation between the two photosynthetic pathways (C3 and C4 plants). In addition, due to the different photosynthetic pathways' environmental controls, the N absorption by C4 plants might be more affected by the competitive pressure of neighboring plants and soil microorganisms than C3 plants along the transect, resulting in the variation in C4 plants' δ 15 N being more sensitive to precipitation [26]. This demonstrates that the responses of plant δ 15 N to environmental changes may also be dependent on the photosynthetic pathways [41]. Although the δ 15 N value of C3 and C4 herbs was correlated with the MAP, the simple relationship between δ 15 N signal and MAP only explained a small amount of the observed variance in the transect. The MAP accounted for just 10.40% and 25.03% of the nitrogen isotope variations in C3 and C4 herbs, respectively (Figure 5a,b). This means that δ 15 N signature of herbs has a limited indicative significance for the variation in precipitation across the study region. Multiple regression analysis (Table 1) also shows that the regression coefficients of the MAP against plant δ 15 N values are only −0.035 (C3 herbs) and −0.053 (C4 herbs), respectively, which again indicates that the influence of precipitation on the δ 15 N of herbs is limited in the study transect. This phenomenon may be related to the fact that there is no obvious precipitation gradient on the study transect. This is because although precipitation is considered one of the key factors to determine plant δ 15 N in arid and semi-arid environments, most of the plant samples in this study were collected along the 400 mm isoline of MAP, and the MAP between the sampling sites varies from 345.0 to 442.8 mm, with an average value of 394.77 ± 24.07 mm. Therefore, there is no significant difference in the effect of precipitation on foliar δ 15 N of herbs due to little difference in precipitation among various sampling sites on the study transect.     [23,24]. The reason for this phenomenon may be that the C3 plant samples used included a large number of woody plants, so it was difficult to distinguish the effects of the photosynthetic pathway on plant δ 15 N values. As for the reason why δ 15 N in C3 grass was more negative in response to increased precipitation than C4 grass in Australian grasslands, Murphy and Bowman did not give a reasonable explanation [41].
This negative relationship between herb δ 15 N and precipitation in this study area is consistent with previous research on local (e.g., Northeast China Transect (NECT), Southern Africa, and Loess Plateau in Northwest China), regional and global scales [4,10,11,14,21,23,26,30,41]. For example, a study conducted across a 1200 km transect of Inner Mongolian grasslands showed that the δ 15 N of two grass species was negatively correlated with MAP [42]. The study by Craine et al. based on over 11000 non-N2-fixing plants worldwide is another example of the pattern of increasing δ 15 N with decreasing MAP [10]. However, the response degree of plant δ 15 N to precipitation varies with different study areas. In the study area, the response of plant δ 15 N to the MAP is significantly greater than that obtained in the Loess Plateau of China (plant roots: −1.1%/100 mm; plant residue: −1.4%/100 mm), NECT (C3 herbs: −1.3%/100 mm; C4 herbs: −1.1%/100 mm) and South Africa (−0.47/100 mm) [4,30].
At present, there are many explanations for the decrease in plant δ 15 N with increasing precipitation. Sutton et al. explained the decrease in plant δ 15 N caused by the increase of precipitation by describing that plants directly absorbed NH in atmospheric precipitation through leaf stomata or epidermis, and the δ 15 N from ammonium nitrogen in precipitation is often more negative than that from nitrate nitrogen, which leads to the reduction in plant δ 15 N [43]. A popular explanation for higher plant δ 15 N in lower-precipitation areas  However, contrary to our findings, Swap et al. and Zhao et al. reported that the relationship between plant δ 15 N and precipitation in Southern Africa and in the Tengger Desert of China was much stronger and steeper in C3 than C4 plant [23,24]. The reason for this phenomenon may be that the C3 plant samples used included a large number of woody plants, so it was difficult to distinguish the effects of the photosynthetic pathway on plant δ 15 N values. As for the reason why δ 15 N in C3 grass was more negative in response to increased precipitation than C4 grass in Australian grasslands, Murphy and Bowman did not give a reasonable explanation [41].
This negative relationship between herb δ 15 N and precipitation in this study area is consistent with previous research on local (e.g., Northeast China Transect (NECT), Southern Africa, and Loess Plateau in Northwest China), regional and global scales [4,10,11,14,21,23,26,30,41]. For example, a study conducted across a 1200 km transect of Inner Mongolian grasslands showed that the δ 15 N of two grass species was negatively correlated with MAP [42]. The study by Craine et al. based on over 11000 non-N2-fixing plants worldwide is another example of the pattern of increasing δ 15 N with decreasing MAP [10]. However, the response degree of plant δ 15 N to precipitation varies with different study areas. In the study area, the response of plant δ 15 N to the MAP is significantly greater than that obtained in the Loess Plateau of China (plant roots: −1.1%/100 mm; plant residue: −1.4%/100 mm), NECT (C3 herbs: −1.3%/100 mm; C4 herbs: −1.1%/100 mm) and South Africa (−0.47/100 mm) [4,30].
At present, there are many explanations for the decrease in plant δ 15 N with increasing precipitation. Sutton et al. explained the decrease in plant δ 15 N caused by the increase of precipitation by describing that plants directly absorbed NH + 4 in atmospheric precipitation through leaf stomata or epidermis, and the δ 15 N from ammonium nitrogen in precipitation is often more negative than that from nitrate nitrogen, which leads to the reduction in plant δ 15 N [43]. A popular explanation for higher plant δ 15 N in lower-precipitation areas is mainly related to greater N losses at drier sites and to 15 N enrichment in soil because of 15 N-depleted N loss through volatilization, denitrification and leaching [18]. In other words, dry sites have a more open N cycle with a greater importance of inputs and outputs compared to humid sites. This is because the acquisition of nitrogen by plants mainly comes from the absorption of soil inorganic nitrogen (ammonium nitrogen and nitrate nitrogen), while the impact of increased precipitation on soil available nitrogen is mainly realized by reducing microbial activity and changing the relative content of ammonium nitrogen and nitrate nitrogen in the soil. Therefore, the nitrogen isotope fractionation occurs during the above process. In general, with the decrease in the aridity index, the utilization efficiency of NO − 3 in soil increases, and more NO − 3 is absorbed by plants and stored in soil nitrogen pool, resulting in an increase in NO − 3 ions and 15 N-depleted nitrogen in soil. Meanwhile, with the increase in soil moisture, soil microbial activity is reduced and the soil nitrification process is inhibited, which causes the availability of soil inorganic nitrogen to decrease and inhibits 15 N enrichment in the soil nitrogen pool, thus resulting in negative δ 15 N values in soils. In the relatively arid area of the FPENC, soil microbial activity increases due to the reduction in annual precipitation and high rates of evaporation, which leads to enhanced ammonification and openness of the nitrogen cycle in the soil. Thus, the ammonium nitrogen in the soil is easy to volatilize on the soil surface, thereby enriching 15 N in the soil nitrogen pool. Because plant δ 15 N was positively correlated with soil δ 15 N in our study (Figure 7), the δ 15 N values of herbs showed a negative correlation with increasing precipitation, but this was weak for C3 herbaceous plants (Figure 5a,b). These paralleled δ 15 N values between plants and soil support the fact that soil N was the dominant N source for the C3 and C4 herbs in the present study, and also indicate an effective internal recycling of nitrogen within the plant-soil system. is mainly related to greater N losses at drier sites and to 15 N enrichment in soil because of 15 N-depleted N loss through volatilization, denitrification and leaching [18]. In other words, dry sites have a more open N cycle with a greater importance of inputs and outputs compared to humid sites. This is because the acquisition of nitrogen by plants mainly comes from the absorption of soil inorganic nitrogen (ammonium nitrogen and nitrate nitrogen), while the impact of increased precipitation on soil available nitrogen is mainly realized by reducing microbial activity and changing the relative content of ammonium nitrogen and nitrate nitrogen in the soil. Therefore, the nitrogen isotope fractionation occurs during the above process. In general, with the decrease in the aridity index, the utilization efficiency of NO in soil increases, and more NO is absorbed by plants and stored in soil nitrogen pool, resulting in an increase in NO ions and 15 N-depleted nitrogen in soil. Meanwhile, with the increase in soil moisture, soil microbial activity is reduced and the soil nitrification process is inhibited, which causes the availability of soil inorganic nitrogen to decrease and inhibits 15 N enrichment in the soil nitrogen pool, thus resulting in negative δ 15 N values in soils. In the relatively arid area of the FPENC, soil microbial activity increases due to the reduction in annual precipitation and high rates of evaporation, which leads to enhanced ammonification and openness of the nitrogen cycle in the soil. Thus, the ammonium nitrogen in the soil is easy to volatilize on the soil surface, thereby enriching 15 N in the soil nitrogen pool. Because plant δ 15 N was positively correlated with soil δ 15 N in our study (Figure 7), the δ 15 N values of herbs showed a negative correlation with increasing precipitation, but this was weak for C3 herbaceous plants (Figure 5a,b). These paralleled δ 15 N values between plants and soil support the fact that soil N was the dominant N source for the C3 and C4 herbs in the present study, and also indicate an effective internal recycling of nitrogen within the plant-soil system. The response of nitrogen isotope discrimination to climate factors is more complex compared with carbon isotope discrimination. Precipitation and nitrogen availability usually play an important role in the nitrogen isotopic discrimination in soil and plants [44]. In our study, the negative effect of MAP on plant δ 15 N values may be due to the reduction in net nitrogen isotope discriminations (∆δ 15 N) in soil. We found that in the FPENC, the ∆δ 15 N values (the δ 15 N difference between soil and plants) decreased with the increase in MAP, but not significantly (Figure 8a,b). This was consistent with the patterns obtained by predecessors in the dryland ecosystem (MAP < 500 mm), but contrary to the patterns obtained in areas with MAP > 800 mm [12,20,44]. This is because a higher MAP is conducive to the loss of nitrogen, which may lead to higher δ 15 N values in wetter soil, but the increased leaching with the increase in MAP can also discriminate against 15   The response of nitrogen isotope discrimination to climate factors is more complex compared with carbon isotope discrimination. Precipitation and nitrogen availability usually play an important role in the nitrogen isotopic discrimination in soil and plants [44]. In our study, the negative effect of MAP on plant δ 15 N values may be due to the reduction in net nitrogen isotope discriminations (∆δ 15 N) in soil. We found that in the FPENC, the ∆δ 15 N values (the δ 15 N difference between soil and plants) decreased with the increase in MAP, but not significantly (Figure 8a,b). This was consistent with the patterns obtained by predecessors in the dryland ecosystem (MAP < 500 mm), but contrary to the patterns obtained in areas with MAP > 800 mm [12,20,44]. This is because a higher MAP is conducive to the loss of nitrogen, which may lead to higher δ 15 N values in wetter soil, but the increased leaching with the increase in MAP can also discriminate against 15 N [18,23].

Response of Foliar δ 15 N values of Herbs to MAT
Temperature is another important environmental factor affecting nitrogen isotope fractionation in plants. Lots of studies have indicated that at regional and global scales, foliar δ 15 N of terrestrial plants tended to increase with increasing temperature. For example, Craine et al. synthesized foliar δ 15 N from global sites and found that foliar δ 15 N increased with increasing MAT [2]. Martinelli et al. reported that the average δ 15 N of plants from tropical regions was 6.5% higher than that from temperate regions (3.7 vs. −2.8%) [16]. In addition, increased plant δ 15 N with rising temperature was also observed along the East African Rift Zone in Ethiopia [15], the Dongling Mountain in Beijing [45], Gongga Mountain [46] and over East China [22]. In this study, however, the variation in foliar δ 15 N of herbs showed a significant negative trend with increasing MAT (Figure 5c-6d and Table  1). The regression analysis of δ 15 N values of C3 and C4 herbs and MAT led to a correlation that was extremely significant (p < 0.001), but the temperature only explained 30.12% (C3 herbs) and 32.48% (C4 herbs) of the observed variation (Figure 5c,d), demonstrating that there was no significant difference in the responses of δ 15 N value of C3 and C4 herbs to temperature. When the MAT increased by 1 °C, the foliar δ 15 N values of C3 and C4 herbs decreased by 0.368% and 0.381%, respectively. In the present study, the variation trend of δ 15 N values of herbs with temperature was opposite to that for all plants at a global scale [10,14], but similar to the results that foliar δ 15 N was negatively correlated with increasing MAT in the Loess Plateau of China [21], the grasslands of the NECT and a climatic gradient transect in South China [20,30,47]. For instance, Feng et al. found that the δ 15 N values of both the C3 and C4 plants declined significantly with increasing MAT. There are three

Response of Foliar δ 15 N Values of Herbs to MAT
Temperature is another important environmental factor affecting nitrogen isotope fractionation in plants. Lots of studies have indicated that at regional and global scales, foliar δ 15 N of terrestrial plants tended to increase with increasing temperature. For example, Craine et al. synthesized foliar δ 15 N from global sites and found that foliar δ 15 N increased with increasing MAT [2]. Martinelli et al. reported that the average δ 15 N of plants from tropical regions was 6.5% higher than that from temperate regions (3.7 vs. −2.8%) [16]. In addition, increased plant δ 15 N with rising temperature was also observed along the East African Rift Zone in Ethiopia [15], the Dongling Mountain in Beijing [45], Gongga Mountain [46] and over East China [22]. In this study, however, the variation in foliar δ 15 N of herbs showed a significant negative trend with increasing MAT (Figure 5c,d and Table 1). The regression analysis of δ 15 N values of C3 and C4 herbs and MAT led to a correlation that was extremely significant (p < 0.001), but the temperature only explained 30.12% (C3 herbs) and 32.48% (C4 herbs) of the observed variation (Figure 5c,d), demonstrating that there was no significant difference in the responses of δ 15 N value of C3 and C4 herbs to temperature. When the MAT increased by 1 • C, the foliar δ 15 N values of C3 and C4 herbs decreased by 0.368% and 0.381%, respectively. In the present study, the variation trend of δ 15 N values of herbs with temperature was opposite to that for all plants at a global scale [10,14], but similar to the results that foliar δ 15 N was negatively correlated with increasing MAT in the Loess Plateau of China [21], the grasslands of the NECT and a climatic gradient transect in South China [20,30,47]. For instance, Feng et al. found that the δ 15 N values of both the C3 and C4 plants declined significantly with increasing MAT. There are three explanations for the positive correlation between plant δ 15 N and temperature. One explanation is that hot environments have a greater proportion of N being lost through fractionating pathways and a more open soil N cycle. More specifically, with the increase in temperature, an increasing fraction of soil N losses are 15 N-depleted forms [4]; the other explanation for this positive relationship may be related to net nitrification rates. The activity of nitrifying bacteria is usually very sensitive to temperature and increases with rising temperature. Thus, in warmer locations, more NH + 4 that is more easily absorbed by plants is converted into NO − 3 which is relatively more difficult to be utilized by plants, resulting in plant δ 15 N values closer to zero [22]. Meanwhile, the net mineralization caused by microorganisms in the soil is enhanced with the increase in temperature. The synergistic effects of the above two effects may make plant δ 15 N and temperature positively correlated. The third explanation is due to the correlation observed at local scale between nitrogen availability and foliar δ 15 N; this positive global relationship between temperature and plant δ 15 N is interpreted as the increased availability of nitrogen in warm environments. Because sites with higher nitrogen availability are more likely to have plants with higher nitrogen concentrations, plant nitrogen concentrations tend to correlate positively with foliar δ 15 N [10]. As for the negative correlation between plant δ 15 N and increasing temperature in the Loess Plateau of China and the NECT, some researchers attributed it to the fact that the significant negative effect of increasing precipitation on plant δ 15 N in the same period of "rain and heat" exceeded the positive effect of rising temperature on plant δ 15 N [21,30]. This means that the negative correlation trend with temperature cannot truly reflect the relationship between plant δ 15 N and temperature in the Loess Plateau and the NECT.
In the present study, we believed that temperature was an important factor affecting the variation in plant δ 15 N, and it was possible that plant δ 15 N values decreased with the increase in temperature. There may be three reasons for this. First, the surveyed transect spans a wide range (the north-south distance is about 1900 km, and the east-west distance about 1500 km) where the MAT was −6.1 to 8.9 • C ( Table 3). The difference between the maximum and minimum temperatures was about 15.0 • C, and there was an obvious temperature gradient from low to high along the northeast to southwest transect. However, the difference in MAP among the various sampling sites on this transect is small, which can basically be regarded as the same (Table 3). Second, by comparing the relationships between temperature/precipitation and longitude/latitude in the transect, it can be observed that both the MAT and the MAP showed a decreasing trend with the increase in latitude and longitude. Among them, the MAT and longitude/latitude were highly correlated. The MAT decreased by 0.43 • C and 0.74 • C for every 1 degree increase in longitude and latitude, respectively, whereas the MAP and longitude/latitude did not show a significant correlation (Table 2). This implies that the influence of temperature with longitude and latitude on plant δ 15 N should be greater than that of precipitation. The multiple linear regression analysis on foliar δ 15 N of herbs and climate factors showed that the absolute value of the regression coefficient of the MAT was much larger than that of the MAP (Table 1), indicating that on this transect, temperature is the key factor affecting plant δ 15 N values. This means that the change in nitrogen isotopes of plants can indicate information regarding temperature change in this region to a certain extent. Thirdly, previous studies also concluded that with the increase in temperature, plant δ 15 N decreased or did not change significantly [41,48]. For instance, on the global scale, foliar δ 15 N decreases with increasing MAT when the temperature is below −0.5 • C [49]. Yi and Yang found that the δ 15 N of alpine meadow plants gradually increased with the decrease in temperature, but it was not significant [50]. Our previous study in the Dongling Mountain of Beijing also showed that when the temperature was less than 3.5 • C, the δ 15 N values of herbs decreased with the increase in temperature, and while the temperature was above the value, the δ 15 N increased with the increase in temperature [45]. As for the increase in foliar δ 15 N with decreasing temperature, it may be that lower temperatures increased the viscosity of soil solution or water in plants and caused soil or plants to be under water stress (i.e., physiological drought), thus increasing plant δ 15 N values. It is worth mentioning that in this study, although the significant negative effect of temperature on leaf δ 15 N occurred, the net N isotope discrimination (∆δ 15 N) in plants and soil increased with the increase in MAT (Figure 8c,d). This showed that soil 15 N-depleted gaseous N losses were greater relative to N input from plant fixation in our study, which also means that N isotope discrimination in plants and soil to temperature variable is very complex.

Sample Collection, Processing and Nitrogen Isotope Determination
From July to August 2020, we established a northeast-southwest transect across the 400 mm isoline of MAP in the APZNC (Figure 9). The transect is approximately 1500 km long and covers approximately 11 • latitude and 18 • longitude (36.92-48.20 • N and 104.20-122.03 • E). Along the entire transect, a total of 22 representative sampling sites were selected 500-1000 m away from major roads and human settlements, without grazing and other anthropogenic disturbances (Figure 9). The longitude, latitude and altitude of each sampling location were measured with GPS (eTrex Venture, Garmin, Kansas City, MO, USA). More details of all sampling sites used in this study are presented in Table 3. At each site, three quadrats with an area of 2.0 m × 2.0 m were randomly selected. Within each quadrat, five to seven mature and healthy individuals of each dominant herb species or eurytopic species were collected, and then combined into one sample of the same species. After air drying in the field, collected plant samples in each site were carefully classified according to their photosynthetic pathway as both C3 and C4, and then stored in envelopes separately. While collecting plant samples, the surface soils (0-10 cm depth) of the corresponding quadrat were also collected systematically by a soil gauge (2.5 cm diameter) in three repetitions. After removing weeds, fine roots, gravel and other sundries, three drill soils were mixed to form one soil sample in each quadrat and put into plastic self-sealing bags. A total of 241 herbaceous plant samples (including 161 C3 plant samples and 80 C4 plant samples) and 66 soil samples were collected in this study.
The plant samples were brought back to the laboratory, dried at 65 • C for 48 h to achieve a constant weight, and then ground to pass through an 80 mesh sieve. Air-dried soil samples were ground to sieve through a 2 mm mesh, and then soaked with 0.5 mol/L hydrochloric acid solution for 24 h to remove carbonates in the soil, and ultimately, the filtered soil extract was used to measure soil δ 15 N. For the prepared plant and soil samples, the δ 15 N, C and N concentrations were determined using a Finnigan MAT-Delte plus XP mass spectrometer (Thermo Finnigan, San Jose, CA, USA) with an automatic continuous-flow Flash EA1112 elemental analyzer (Thermo Finnigan, San Jose, CA, USA), and then the mean value per sampling site was calculated. The nitrogen isotopic composition is defined in per mil (%) relative to atmospheric N 2 and expressed as: where R sample and R standard represent the 15 N/ 14 N ratios of the sample and standard, respectively. The measurement accuracy of nitrogen isotopes was ± 0.3%.  Figure 9. The locations of the sampling sites in the study area. The Arabic numerals in Figure 9 are the numbers of sampling sites, which correspond to the serial numbers in Table 3. Grassland Type Species Names Figure 9. The locations of the sampling sites in the study area. The Arabic numerals in Figure 9 are the numbers of sampling sites, which correspond to the serial numbers in Table 3.

Meteorological Data of Sample Sites
Two main climatic variables (MAT and MAP) were used to explore the climatic controls on response patterns of δ 15 N in C3 and C4 herbs. The MAT and MAP were obtained from local weather stations of the China Meteorological Data Service Centre and Chinese Natural Resources Database. Simultaneously, the aridity index (AI) of each sampling site was calculated by Equation (2).
where PT 0 is the potential evapotranspiration (mm), R 0 is the annual precipitation (mm) and T is the monthly average temperature (If the monthly average temperature exceeds 30 • C, it shall be calculated as 30 • C; when the monthly average temperature is lower than zero • C, it is calculated as 0 • C). The above climatic data were the average of observed data during the 30 years from 1990 to 2020.

Statistical Analysis
Ordinary least squares regression (OLSR) was used to examine the responses of plant δ 15 N to climatic factors as well as to soil δ 15 N. In the process of analysis, only non-N 2fixing herbaceous species were selected for this study. Meanwhile, ten samples of five annual C3 herbs were excluded in order to eliminate the influence of leaf age on plant δ 15 N values. In other words, the C3 plant samples involved in the analysis were all biennial or perennial herbs. Considering the correlation between precipitation and temperature in the study area, partial correlation and multiple linear regression analyses based on the Akaike information criterion (AIC) were used to explore the relationships of plant δ 15 N against climatic variables such as temperature and precipitation, so as to distinguish the effects of temperature and precipitation on plant δ 15 N. In addition, one-way ANOVA (analysis of variance) was performed to test whether there was a significant difference in the average δ 15 N values between C3 and C4 herbs. All statistical analyses were conducted with SPSS 13.0 software (SPSS Inc., Chicago, IL, USA).

Conclusions
Through the investigation of the foliar δ 15 N of herbs in the APZNC and the responses of plant δ 15 N to the MAP and MAT, the following conclusions were preliminarily drawn: (1) The C3 and C4 herbs in the APZNC exhibited an obvious difference in their δ 15 N signatures. The δ 15 N values of all herbs investigated varied from −5.5% to 15.25%. Among them, the δ 15 N value range of C3 herbs (−5.5~15.00%) was wider than that of C4 herbs (−2.17~15.25%), but its average value (3.27%) was significantly lower than that of C4 herbs (5.55%). (2) From northeast to southwest along the transect, the δ 15 N of herbs decreased linearly and significantly with the increase in precipitation, which is consistent with previous related studies. However, the ability of MAP to explain the nitrogen isotope changes in C3 and C4 herbs was weak, only 10.40% (C3 herbs) and 25.03% (C4 herbs), respectively, indicating that the influence of precipitation on plant δ 15 N is limited in this transect. (3) In the transect, the δ 15 N of both C3 and C4 herbs decreased significantly with the increase in MAT. The δ 15 N of C3 and C4 herbs decreased by 0.368‰ and 0.381‰, respectively, for every 1 • C increase in MAT. There was no significant difference in the δ 15 N responses of C3 and C4 herbs to temperature. However, the interpretation ability of temperature to the change in δ 15 N of C3 herbs was significantly higher than that of precipitation, which suggests that temperature is a key factor affecting foliar δ 15 N of herbs in this transect. The above findings increased our knowledge of the δ 15 N signatures of C3 and C4 herbs and their responses to climate change, which could facilitate global change researchers to use plant δ 15 N as a proxy to explain the process of the nitrogen cycle and gain an insight into climate dynamics of the past. Since the sampling area was selected along the 400 mm isoline, the range of tested precipitation is very narrow. In addition, the δ 15 N values of nitrate nitrogen and ammonium nitrogen in the soils were not measured in this study. Therefore, future work is needed across a larger scale with a wide range of precipitation. Moreover, it is necessary to study the effects of different nitrogen sources on plant δ 15 N.
Author Contributions: Sampling, X.L., Y.L. and Y.S.; methodology, X.L. and Y.Z.; data processing, Q.S. and Y.L.; writing-original draft preparation, X.L. and Y.L.; writing-review and editing, T.F.; map making, Q.S. and T.F.; project administration, X.L., Q.S. and Y.Z.; funding acquisition, X.L., Y.Z. and T.F. All authors have read and agreed to the published version of the manuscript. Data Availability Statement: The data will be provided by the corresponding author upon request.