Nitrogen (N) is an essential nutrient for plant growth and development, and many terrestrial ecosystems are adapted to conditions of low N availability (Bobbink et al., 1998). Since the agricultural and industrial revolution, atmospheric deposition of biologically reactive N has increased drastically due to N fertilization and combustion of fossil fuels across the globe (Galloway et al., 2008; Canfield et al., 2010; Sutton and Bleeker, 2013), resulting in a large impact on community composition and function of ecosystems (De Schrijver et al., 2008; Cardinale et al., 2012; H. Yang et al., 2012). Elevated atmospheric N deposition generally has positive effects on productivity for the N limited ecosystems (Smith et al., 1999; Galloway et al., 2008), while it imposes a great threat to biodiversity of the terrestrial ecosystems (Stevens et al., 2004; Clark and Tilman, 2008; Bobbink et al., 2010; Jiang et al., 2010; Kim et al., 2011). Both large-scale field survey and manipulated experiments to simulate N deposition have shown that N deposition has driven significant reductions in plant species richness in different grassland ecosystems (Stevens et al., 2004; Suding et al., 2005; Bai et al., 2010; Clark and Tilman, 2008, Dupre et al., 2010; Van Den Berg et al., 2011). Along with the time of N deposition or with the increase of N addition rate, the decline in species richness by N deposition consequently results in changes of community composition and reduction in ecosystem productivity (Isbell et al., 2013).

Several hypotheses have been proposed to explain the N deposition-induced species loss in grassland ecosystems, such as NH4+ toxicity to plants (van den Berg et al., 2005; Stevens et al., 2006; Zhang et al., 2014), soil acidification (van der Putten et al., 2013), mobilization of toxic metals in soils (Bowman et al., 2008; Horswill et al., 2008; Stevens et al., 2009; Chen et al., 2013) and changes in soil microbial activity and biodiversity (Dean et al., 2014). Competitive exclusion driven by enhanced resource uptake by dominant species and pre-emption of light or space has been widely invoked as a key mechanism for species loss under elevated N regimes (Clark and Tilman, 2008; Hautier et al., 2009; Suding et al., 2005; Borer et al., 2014). For instance, it has been suggested that chronic N deposition shifts grassland towards grass-dominated vegetation due to higher productivity of grasses at elevated N concentrations, which are thought to outcompete forbs and shrubs (Heil and Diemont, 1983; Bobbink et al., 1998; Stevens et al., 2006). However, N addition-induced reductions in plant biodiversity cannot simply be explained by competitive exclusion, because many species had already disappeared before grasses became dominant (Houdijk et al., 1993) and fertilization also reduces plant biodiversity of grassland even when light is not limiting (Dickson and Foster, 2011). Although extensive research has demonstrated that N deposition reduces biodiversity, the primary mechanism underlying the N deposition-induced changes in community composition remains largely unknown (Stevens et al., 2006; Phoenix et al., 2003; Bowman et al., 2008; Clark and Tilman, 2008; Hautier et al., 2009; Suding et al., 2005; Borer et al., 2014).

N deposition often concurs with soil acidification (Stevens et al., 2004; Bobblink et al., 2010; Fang et al., 2012; Horswill et al., 2008; Y. Yang et al., 2012). Soil acidification subsequently mobilizes some metal ions, thus rendering phytotoxicity to plants at high concentrations (Kochian, 1995; Marschner, 1995). For instance, release of toxic aluminum (Al3+) ions due to soil acidification has been suggested to be a driving force for N deposition-induced species loss in grasslands (Carnol et al., 1997; Horswill et al., 2008; Chen et al., 2013). In addition to Al3+, homeostasis of other ions in soil is also closely determined by soil pH, such that reduction in soil pH would enhance release of those metal ions of Fe3+, Mn2+ and Cu2+ (Marschner, 1995; Bowman et al., 2008). The involvement of soil acidification-mediated processes in species loss under elevated N deposition has been extensively evaluated in acidic grasslands (Stevens et al., 2006; Bowman et al., 2008; Horswill et al., 2008). Whether this mechanism is also responsible for N deposition-induced changes in plant biodiversity in other types of grassland remains largely unknown. A major difference between acidic grasslands and temperate steppe used in the present study lies in their basic properties of soils, such as soil pH, ion contents and acid buffering systems. In acidic grasslands, soil pH is usually < 5.0, and availabilities of metal ions, such as Al3+, Fe3+, Mn2+ are high compared to those in the alkaline soils, and acid buffering is mainly dependent on aluminium, leading to lower acid buffering capacity (Bowman et al., 2008). However, soils in neutral or alkaline grasslands have more base cations, higher acid buffering capacity and low availabilities of metal ions (Al3+, Fe3+, Mn2+). In addition to the differences in soil traits, plants grown in acidic and alkaline grasslands may also have evolved adaptive strategies to their edaphic conditions. Plants in the alkaline temperate steppe would be exposed to high levels of metal concentrations due to N deposition-driven soil acidification, rendering them metal toxicity. Therefore, plants in the alkaline grasslands and acid grasslands may differ in their sensitivity to N deposition-induced changes in soil traits

Inner Mongolia grassland is an important part of widely distributed grasslands across the Eurasian Steppe with typical calcareous soil distinguished by high pH and buffering capacity due to abundant base cations (Chen et al., 2013). These differences in soil traits between the temperate steppes and acid grasslands may render the two types of grasslands differing in their sensitivity to N deposition. N deposition rate in China has increased dramatically in recent decades (Liu et al., 2013; Jia et al., 2014), thus imposing great threats to plant biodiversity in grassland ecosystems. Moreover, a significant soil acidification in grasslands across northern China over the past 2 decades has been reported (Y. Yang et al., 2012). In contrast to acidic grasslands, few studies have investigated the role of soil acidification-driven metal mobilization in species richness in calcareous and alkaline grasslands under conditions of elevated N deposition. To evaluate the role of soil-mediated chemical processes in N deposition-induced changes in species richness, the chronic effects of N addition, which simulates N deposition, on soil pH, nutrient availability and species composition were investigated in a temperate steppe of Inner Mongolia.

Nitrogen deposition has multiple impacts on grassland ecosystems, including changes in productivity, reductions in species richness and soil acidification (Smith et al., 1999; Stevens et al., 2004; Galloway et al., 2008; Clark and Tilman, 2008; Bobbink et al., 2010). Elevated N deposition resulting from human activities in the past decades has negative impacts on growth and development of certain plant species, leading to reduction in plant diversity (Isbell et al., 2013; Maskell et al., 2010; Stevens et al., 2006). To evaluate the effects of N deposition on temperate grassland ecosystems, long-term N fertilization experiments were conducted in Inner Mongolia steppes by applying urea. The applied urea can be hydrolyzed to ammonia/ammonium, by the enzyme urease, and ammonium is further converted into nitrate by ammonia oxidizing bacteria and ammonia oxidizing archaea, leading to an increase in inorganic N in soils and concurrent reduction in soil pH (Zhang L. M. et al., 2012). Previous studies showed that the application of urea led to significant increases in soil nitrate concentrations and soil acidification (Fang et al., 2012), which are consistent with this proposition. Despite potential differences in natural N deposition and application of urea in terms of soil acidification and enrichment of soil N, our N addition experiments can simulate the natural N deposition. In the present study, we found that a low and moderate rate of N addition for consecutive 9 years led to an enhanced total AGB (Fig. 1a). More specifically, we found that AGB of graminoids and forbs displayed different responses to N addition, such that AGB of graminoids and forbs was increased and relatively unchanged by moderate N addition, respectively (Fig. 1c). Species richness of graminoids and forbs also differed in their responses to N addition. N addition significantly reduced forb species richness, while graminoid species richness was relatively unchanged in response to the N addition (Fig. 1d). These findings that forbs were more sensitive to N deposition than graminoids in terms of species richness are consistent with the results obtained in semi-natural European grasslands (Stevens et al., 2006).

Species loss induced by N deposition on grasslands has been suggested to result from competition due to increased growth of graminoids in response to N enrichment (Stevens et al., 2006). Although a moderate rate of N addition, i.e. 2 and 4 g m-2 yr-1, stimulated grass growth, the total species richness and forb species richness under these N addition rates were relatively constant (Fig. 1). Moreover, at higher rates of N addition (16 and 32 g m-2 yr-1), AGB of graminoids was not enhanced, but species richness of forbs was dramatically reduced (Fig. 1). These results may suggest that loss of forbs is not simply caused by competitive exclusion driven by increased growth of graminoids, rather these findings may highlight the involvement of other processes associated with N addition in inhibition of forb growth.

Enhanced N deposition may decrease plant diversity by enrichment of nitrogen nutrient (van den Berg et al., 2005; Stevens et al., 2006; Zhang et al., 2014). In our study, N addition led to significant increases in inorganic N in soils (Fig. 3a). Forb species richness and AGB was negatively correlated to soil inorganic-N concentration (Table A2). Moreover, inorganic N in soil accounted for 23.59 % of the variation in forb AGB (Table 2). These results suggest that an increase in N availability due to N addition may contribute to N deposition-induced loss of forb species. Stevens et al. (2006) and Zhang et al. (2014) demonstrated that species richness is negatively correlated with soil NH4+–N concentrations in both acidic grasslands and alkaline grasslands. Processes associated with N transformation in soils, including mineralization and nitrification, depend on soil pH, which determine homeostasis of NH4+–N and NO3–N (Dorland et al., 2004). Although it has been reported that species from acidic and alkaline soils usually prefer different forms of nitrogen (Falkengren-Grerup and Lakkenborg-Kristensen, 1994; van den Berg et al., 2005), both high NO3-–N concentration and high NH4+–N concentration can suppress root elongation (Britto and Kronzucher, 2002; Tian et al., 2005, 2009; Zhao et al., 2007). Therefore, it is necessary to further dynamically monitor changes in N forms in soils after N addition and to evaluate the different response of forbs and graminoids to different N forms. In addition to enrichment of N, an increase in soil P availability has been implicated in the reduction of species richness of grasslands (Ceulemans et al., 2013). In the present study, N addition significantly increased the Olsen-P concentration in soils (Fig. 3). The increase in P availability may result from N addition-induced soil acidification because P is mainly precipitated as calcium phosphate in calcareous and alkalinous soil, its solubility would be enhanced by reduced soil pH. Although Olsen-P concentration in soils exhibited negative correlation with forb species richness (Table A3), multiple regression showed that P availability did not contribute to the N-induced changes in forb species richness and biomass (Table 2). These results discount the involvement of P availability in the N-induced changes in species richness.

Soil acidification often concurs with N deposition due to the formation of hydrogen ions during ammonia oxidation (Guo et al., 2010; Y. Yang et al., 2012). Numerous studies across N deposition gradients (Maskell et al., 2010; Stevens et al., 2004) and field fertilization experiments (Bowman et al., 2008; Lan and Bai, 2012; Zhang et al., 2014) have demonstrated that N deposition leads to soil acidification. In the present study, we found a significant soil acidification by 9-year N addition in the calcareous temperate grassland (Fig. 2). A positive correlation between soil pH and the species richness was found in this study (Table A2). However, Chytrý et al. (2007) reported that in tundra and forest with low soil pH, species richness is increased with soil pH, while in steppe with soil pH > 6.0, the species richness appears to be negatively dependent on soil pH. The differences between our results and those of Chytrý et al. (2007) may be accounted for by the differences in soil traits because the relationship between species richness and soil pH is dependent upon vegetation types, soil traits and climatic conditions (Chytrý et al., 2007). Soil pH in our study was reduced from 6.82 to 5.37 by the N addition (Fig. 2). This pH range is comparable to that of forest surveyed by Chytrý et al. (2007). In this range of soil pH, they also discovered that species richness is positively correlated with soil pH (Chytrý et al., 2007). Species richness showed positive correlation between soil pH suggests that soil pH is an important factor in determination of species richness in the Inner Mongolia steppe.

Soil acidification would disturb ion homeostasis in soil, including depletion of base cations and mobilization of metal cations (Bowman et al., 2008; Horswill et al., 2008). Ca2+, Mg2+ and K+ are dominant base cations in calcareous soils. Depletion of these base cations in soils by N addition (Fig. 3b) would render the soil less capable of buffering acid. The insignificant correlation between K+ concentrations and soil pH, and N addition rates rules out the possibility that soil K+ may contribute to the decline in plant diversity (Table A1). Multiple regression showed that changes in Ca2+ and Mg2+ concentrations reflected by F2 contributed little to AGB and species richness of forbs, suggesting that patterns in species richness of forbs are not driven by depletion of Ca2+ and Mg2+ induced by N addition.

In addition to depletion of base cations, soil acidification can release some metal ions by increasing their solubility. N addition markedly enhanced concentrations of Mn2+, Fe3+, Cu2+ and Al3+ in soils (Fig. 3c), and changes in these metal ions explained more variation on forb species richness than inorganic-N (Table 2). These results suggest that the release of metal cations is a main driving force for N addition-evoked loss of forb species. In contrast to Mn2+, Fe3+ and Cu2+, an increase in soil Al3+ concentration was not detected with N addition rate < 32 g m-2 yr-1 (Fig. 3c), while loss of forb species had already occurred at moderate N addition rate (8 g m-2 yr-1) (Fig. 1d). These results may imply that soil Al3+ is unlikely to be a key driving factor for species loss evoked by N deposition in our experimental systems. Several studies have demonstrated that Al toxicity is involved in N deposition-induced species loss in acidic grasslands with soil pH < 5 (Stevens et al., 2009). In our studies, soil pH was greater than 5 even under the highest N addition rates (Fig. 2). Moreover, changes in Al3+ concentration had lower partial correlation coefficient (p-R2) with soil pH (Table A3). Given that Al phytoxcity normally occurs at soil pH < 5 (Tyler, 1996), the contribution of Al toxicity to species loss can be discounted in our studies. Similar to results reported by Bowman et al. (2008), we found that N addition led to a substantial increase in soil Fe3+ concentration. Because forbs can only take up Fe2+ after reduction of Fe3+ to Fe2+ by ferric chealate reductases in roots (Marschner, 1995), the N addition-induced increase in soil Fe3+ concentration would contribute little to the loss of forb species in the present study. Based on the results of linear regression analyses, compared with Cu2+, Mn2+ exhibited the closest correlation with soil pH and Mn2+ concentrations were most greatly affected by N addition (Fig. 3, Table A1). Therefore, mobilization of Mn2+ due to soil acidification induced by N addition is expected to be a critical process responsible for forbs loss in the Inner Mongolia steppe under elevated N deposition, because availability of ions in soil affects nutrient uptake of plants (Marschner, 1995). Forb species would be more prone to accumulate Mn than graminoid species because of their intrinsic differences in biochemical pathways to regulate metal transport (Marschner, 1995). Therefore, further studies to evaluate the effects of N addition on accumulation of metals in general and Mn in particular by forbs and graminoids would provide a biochemical explanation for loss of forb species in the steppe under elevated N deposition.

We demonstrate that N addition reduced species richness, acidified soil and disturbed nutrient homeostasis in soil in an Inner Mongolia steppe. We further reveal that decline in species richness by N addition was mainly accounted for by loss of forb species as forbs were more sensitive to N addition than graminoids. Our findings also show that N addition resulted in an increase in inorganic-N concentration, depletion of base cations (Ca2+ and Mg2+) and mobilization of Mn2+ ions. Nitrogen availability and release of Mn2+ ions were involved in changes of biomass and diversity in the temperate steppe. These findings highlight that soil acidification-mediated Mn2+ mobilization is a key factor to drive decline in the species richness of forbs under elevated N addition in the alkaline, calcareous grasslands in northern China.