The two studied grasslands are located in the Austrian Central Alps near Neustift, Stubai Valley (47∘07′ N, 11∘19′ E) and are characterized by a mean annual temperature of 3∘C and a mean annual precipitation of about 1100 mm. Drought simulations were conducted on a typical extensively managed mountain meadow (to be referred to as “meadow”, 1850 m a.s.l.) and on an abandoned meadow site (to be referred to as “abandoned site” 1900 m a.s.l.). At both sites, soils were characterized as dystric cambisols (FAO classification) with a pH in the uppermost 10 cm soil depth of 5.5 (determined in CaCl2). At the meadow total aboveground plant biomass is cut and harvested once a year; parts of the meadow are slightly grazed by cattle in spring and autumn, and fertilized with manure every two to three years (Bahn et al., 2006). The meadow is characterized by high plant primary production (Schmitt et al., 2010), by higher soil respiration rates (Bahn et al., 2008), as well as by lower soil organic matter (SOM), total carbon (Ctot) and nitrogen (Ntot) contents compared to the abandoned site (Table 1, and Meyer et al., 2012). At the abandoned site all management activities were terminated in 1983. The dominant plant community was determined as Trisetetum flavescentis at the meadow, and as Seslerio-Cariecetum at the abandoned site (Schmitt et al., 2010).

At both the meadow and the abandoned site, four drought and corresponding control plots were established in spring 2011 in a paired plot design. Drought was simulated by installing rain-out shelters, covering an area of 3.0 × 3.5 m, for 10 weeks starting on 31 May 2011. The shelters were equipped with light- and UV-B-permeable plastic foil (UV-B Window, Foiltec GmbH, Germany, light-permeability ca. 95 %, UV-B permeability > 70 %) to exclude any precipitation. To maintain the traditional land-management, the total area of the meadow was mown (i.e. aboveground biomass was cut and removed) during the drought period (2 August 2011), while there was no land-management activity at the abandoned site. After 10 weeks of drought (10 August 2011) all plots received previously collected rainwater over a time period of 3 min to simulate a short heavy rainfall event of 20 mm; subsequently rain-out shelters were removed and all plots were again exposed to natural precipitation.

Soils were sampled 1 week before the onset of drought simulation, then every 2 to 3 weeks during drought, 1 day and 7 weeks after rewetting (Fig. 1). For each sample two soil collars (5 × 7 cm, 10 cm depth) were pooled after the uppermost litter layer had been removed. Then soil collars were homogenized and sieved to 2 mm. Aliquots of soil were immediately frozen at -80 ∘C for molecular analyses; the remaining samples were stored at 4 ∘C until further processing.

For both sites maximum soil water holding capacity (WHCmax) was determined by adding excess amounts of deionized water to aliquots of soil samples and leaving them for 48 h over water to allow maximum saturation. Then 2 g of water-saturated soils were dried for 48 h at 60 ∘C. Soil water content (SWC) was measured gravimetrically by drying of 5 g of fresh soil for two days at 60 ∘C in a drying oven and calculated relative to WHCmax. Total carbon (Ctot) and N (Ntot) content were determined from dried and ground soil samples by EA-IRMS (EA 1110, CE Instruments, Italy, coupled to a Finnigan MAT Delta Plus IRMS, Thermo Fisher Scientific, MA, USA). Total extractable N was determined from K2SO4 extracts (2 g of soil were extracted with 20 mL 0.5 M K2SO4) using a TOC / TN analyser (TOC-V CPH E200V/TNM-122V; Shimadzu, Austria). NH4+ was measured photometrically from K2SO4 extracts using a modified indophenol reaction method (Kandeler and Gerber, 1988). NO3- was determined from water extracts (2 g of fresh soil were extracted with 20 mL of MilliQ water) by chemically suppressed ion-chromatography (DX500, Dionex, Austria) on a Dionex AS11 column. Extractable organic nitrogen (EON) was calculated by subtracting inorganic (NH4+ and NO3-) from total extractable N.

Potential gross NH4+ and NO3- transformation rates were determined using 15N pool dilution techniques described by Kaiser et al. (2011). For determination of microbial gross N mineralization and NH4+ immobilization, fresh aliquots of soil (2 g in duplicates) received 0.5 mL (NH4)2SO4 (0.125 mM; 10 atom % 15N). For determination of gross nitrification and microbial NO3- uptake (NO3- immobilization), 2 g of fresh soil samples (in duplicates) received 0.5 mL KNO3 (0.25 mM, 10 atom % 15N). Labelled samples were incubated for 4 and 24 h at room temperature and finally extracted with 20 mL 2 M KCl. Both control and drought-treated soils received solute 15N label, which could have altered the conditions, specifically in drought treated soils (Chen et al., 2011). Therefore, the reported rates should be considered as the N-turnover potential of the soil microbial community.

Potential gross N mineralization and NH4+ immobilization rates were determined by microdiffusion of NH3 from KCl-extracts into acid traps, which were analysed for nitrogen concentrations and atom-percent excess of 15N by EA-IRMS (EA 1110, CE Instruments, Italy, coupled to a Finnigan MAT Delta Plus IRMS, Thermo Fisher Scientific, MA, USA). For potential gross nitrification and NO3- immobilization rates NH3 was removed from the extracts, before converting NO3- to NH3 by adding Devarda's Alloy. Again, NH3 was trapped and analysed for N concentration and atom-percent excess of 15N. Potential gross N mineralization, gross NH4+ immobilization, as well as potential gross nitrification and gross NO3- immobilization rates were calculated as described in detail by Wanek et al. (2010).

DNA was extracted from 0.35 g of soil using the FastDNA^® SPIN Kit for Soil (MP Biomedicals, CA, USA) and the Precellys24 Instrument (Bertin Technologies, France). After extraction, the DNA was tested in quantity and quality with a spectrophotometer (Nanodrop, PeqLab, Germany) and stored at -20 ∘C until further processing.

The abundances of the bacterial and archaeal ammonia-monooxygenase gene (amoA) served as proxy for ammonia-oxidizers and were detected by quantitative Real-Time PCR, which was carried out on a 7300 Real-Time PCR System (Applied Biosystems, Germany) using SYBR green as fluorescent dye. The PCR was performed in 96-well plates (Applied Biosystems, Germany) for all investigated genes. The reaction mixes were performed according to Töwe et al. (2010); the thermal profiles of the PCRs are given in Table 2. In a preliminary test dilution series of the DNA extracts were tested to avoid inhibition of PCR, resulting in an optimal dilution of 1:128 for all samples (2–5 ng DNA µL-1 per sample). Serial plasmid dilutions of the respective functional genes ranging from 106 to 101 gene copies µL-1 were used as standards for the determination of the gene abundances of each sample (Table 2). To confirm the specificity of the amplicons after each PCR run, a melting curve and a 2 % agarose gel stained with ethidium bromide were conducted. The efficiencies (Eff) of the amplification were calculated from the standard curve with the formula Eff = [10(-1/slope)-1]×100 % and resulted in the following values: amoA of ammonia-oxidizing archaea (AOA) 83.4 % to 91.8 % (R2 = 0.997) and amoA of ammonia-oxidizing bacteria (AOB) 94.8 to 95.4 % (R2 = 0.999).

To determine site-specific differences, all available parameters were compared by repeated-measures ANOVA using only the controls of both sites. As we found significant differences between sites, we subsequently tested for effects of drought by comparing control samples from each study site with drought plots over the course of the experiment by repeated-measures ANOVA, using the plot number as within-factor. To test for differences between drought and control samples at single sampling time points we used paired t-tests with Bonferroni-adjusted levels of significance. All data were tested for normality by Shapiro–Wilk test and for homoscedasticity by Levene's test. If data did not meet ANOVA assumptions they were log-transformed or rank-normalized. All data, including SWC, Ctot, Ntot, all N pools and potential N-turnover data, as well as ammonia-oxidizer gene abundances were subjected to principal component analyses (PCA), after they were transformed to meet PCA assumptions. To determine differences between sites and drought on the distribution of the samplings along the PC-axes, we conducted a two-way ANOVA. All statistical analyses were performed using R 2.15.2 (R Core Team, 2012).

During drought simulation, in total 358 mm of precipitation were excluded, equalling one-third of mean annual precipitation. This led to a significant decrease of SWC to 16.9 % (±1.4 %) at the meadow, and to 21.2 % (±3.5 %) at the abandoned site (Table 1, Fig. 1). However, the initial WHCmax was significantly lower at the meadow than at the abandoned site (Table 1), thus the amount of water remaining in meadow soils was lower during drought as compared to the abandoned site. At both sites, rewetting had no immediate effects on the SWC in drought-treated plots, but 7 weeks after the end of drought the SWC levels had approximated those of controls (Fig. 1).

At both sites drought had no significant effect on the N pools (Table 2, Fig. 2) when concentrations were calculated per gram dry soil, although the meadow was characterized by significantly lower EON and NH4+, but higher NO3- concentrations compared to the abandoned site (Table 1, Fig. 2). However, when N concentrations were calculated per g soil water after 5 weeks of drought, EON and NH4+ concentrations were significantly increased at both sites, and remained on a higher level until 1 day after rewetting (Table 1, Fig. 3). In particular at the meadow, drought induced fivefold increases of NH4+ concentrations, while at the abandoned site NH4+concentrations at most tripled. In contrast, NO3- concentrations in the soil solution were not affected by drought, but strongly increased 1 day after rewetting at the meadow, while at the abandoned site no clear effect of drought was detected (Table 1, Fig. 3). At both sites 7 weeks after the end of drought EOC and NH4+ concentrations were similar in drought and control plots – only in soils of the meadow were NO3- concentrations in drought-treated soils significantly increased (Table 1, Fig. 3).

Potential gross rates of microbial N mineralization and NH4+ immobilization were similar at both sites (Table 1, Fig. 4). However, only at the meadow drought significantly increased potential gross NH4+ immobilization rates, which consequently decreased the mean residence time (MRT) of NH4+ (per g dw) from 27.4 ± 6.8 h to 14.4 ± 1.3 h. At the abandoned site, in contrast, potential gross N mineralization and immobilization, as well as the MRT of NH4+ (50.8 ± 10.5 h) were not affected by drought (Table 3, Fig. 4). At neither site potential N mineralization, nor NH4+ immobilization responded to rewetting.

Potential gross rates of nitrification and microbial NO3- immobilization were significantly lower and the NO3- pool turned over much slower (30.3 ± 9.9 h) at the meadow than at the abandoned site (2.2 ± 0.7 h Table 1, Supplement Fig. S1). At both sites potential nitrification and microbial NO3- immobilization rates were influenced by sampling time, indicating a strong seasonal variability (Table 1, Fig. 4). At the meadow drought affected neither potential nitrification, nor potential NO3- immobilization rates. At the abandoned site, in contrast, drought significantly altered both potential nitrification and NO3- immobilization rates, which showed less variance compared to controls (Table 3, Fig. 4). One day after rewetting at both sites potential nitrification and NO3- immobilization rates slightly increased in drought-treated plots, but seven weeks after rewetting, the rates were again similar to the controls (Fig. 4).

The abundance of archaeal amoA genes (as a proxy for AOA) was similar at both sites, ranging from 8.1 ± 105 to 3.2 ± 106 copies g-1 DW soil at the meadow and from 5.2 ± 105 to 3.7 ± 106 copies g-1 DW soil at the abandoned site. At the meadow bacterial amoA gene copies (as a proxy for AOB), ranging from 2.5 ± 105 to 1.7 ± 106 copies g-1 DW soil, were as abundant as archaeal amoA gene copies, thus AOA : AOB ratios ranged between 1 and 10. At the abandoned site, in contrast, AOB numbers ranged from 8.0 ± 104 to 7.8 ± 105 copies g-1 DW soil, which was significantly lower, consequently AOA : AOB ratios were significantly higher (ranging between 10 and 100) than at the meadow (Table 1, Fig. 5).

At the meadow drought significantly decreased archaeal amoA gene copy numbers, while the abundance of bacterial amoA remained unaffected, thus the ratio of AOA : AOB significantly decreased compared to the controls (Table 3, Fig. 5). At the abandoned site, in contrast, drought affected neither AOA, nor AOB abundances.

To summarize and illustrate drought effects in relation to site-specific differences, a PCA was conducted, including all studied parameters. In total, three factors with eigenvalues > 1 were identified and accounted for 68.9 % of the total variance. In a biplot showing the first two factors (PC1 and PC2, accounting in total for 55.1 % of the variance) on PC1 samples were separated (34.9 %) according to sampling site (two-way ANOVA with factor loadings: (F(1)=102.4; p<0.001) and drought treatment (F(1)=5.9; p<0.05, interaction: ns). PC2 explained 20.1 % and primarily displayed drought effects (F(1)=9.1; p=<0.01), but was not affected by site (F(1)=0.9; p= ns; interaction: ns). The main parameters responsible for the separation along PC1 were higher EON and NH4+ concentrations, as well as higher potential gross nitrification and gross NO3- immobilization rates at the abandoned site, which were in contrast to higher abundances of AOB and NO3- concentrations that were dominating at the meadow (Fig. 6). This pattern was supported by positive correlations between EON and NH4+ concentrations in controls of both the meadow (R=0.59, p<0.001) and the abandoned site (R=0.57, p<0.01). Moreover, in controls of the meadow NO3- correlated negatively with EON concentrations (R=-0.67, p<0.001), but positively with AOB abundances (R=0.42, p<0.05). Only in controls of the abandoned site EON correlated with gross N mineralization (R=0.60, p<0.001) and with SWC (R=0.60, p<0.001). PC2 separated the drought treatment from the controls, although this effect was stronger for the meadow than for the abandoned site. SWC was distributed along PC2 opposing the vectors for potential gross N mineralization and gross NH4+ immobilization (Fig. 6). Specifically in soil samples from the meadow subjected to drought the potential gross NH4+ immobilization rates correlated negatively with SWC (R=-0.62, p<0.001).

In our study the potential of the microbial community to mineralize organic nitrogen to ammonium was not affected by drought irrespective of site-specific differences related to land management. Thus, this comparably “broad” soil process remained stable and could show to indicate that N mineralization may have been dominated by drought-tolerant microbial generalists like fungi (Allison and Martiny, 2008; Schimel and Schaeffer, 2012). Only in soils of the managed meadow did potential gross rates of microbial NH4+ immobilization and therefore NH4+ turnover increase during drought. This points to a potentially increased microbial N demand to facilitate the synthesis of nitrogenous osmolytes to balance the osmotic potential between the soil solution and microbial cells (Schimel et al., 2007). At the abandoned site, however, the microbial NH4+ uptake potential remained unaffected by drought, which suggests a high stability of this process, or a microbial community that might use other than N-containing substrates for osmolyte production (Schimel et al., 2007).

As microbial nitrification appears to be a more “narrow” process than microbial N mineralization (Schimel and Schaeffer, 2012), we expected nitrification rates to respond sensitively to dry conditions. In the presented study, however, responses of potential nitrification to drought were diverging between the studied sites. In accordance with other studies, where grassland and heathland sites had been subjected to drought (Hartmann et al., 2013; Larsen et al., 2011), the microbial nitrification potential was not affected by drought at the managed meadow. At the abandoned site, in turn, drought reduced the dynamics of potential nitrification and NO3- immobilization, which was also shown, e.g. for forest soils (Chen et al., 2011), or grassland under oak forest (Stark and Firestone, 1996). Thus, responses of potential nitrification rates to drought seem to be strongly context dependent and may not be generalized.

Drought distinctly affected bacterial and archaeal amoA gene copy numbers, which were used as proxy for the abundance of AOB and AOA. As expected, AOB were stable and not affected by drought at both sites. The observed decrease of AOA in response to drought at the managed meadow in our study was similar to results reported by Thion and Prosser (2014), where the observed decrease of archaeal amoA abundance during drought was likely related to strong increases of NH4+ during drought. In our study soil NH4+ concentrations calculated per gram dry soil were not affected during drought at either site, similar to the study of Hartmann et al. (2013) in subalpine grassland soils. When NH4+ concentrations were calculated per g soil water, however, fivefold increases at the meadow and up to threefold increases at the abandoned site were observed. Thus, it is likely that the stronger increase of NH4+ in the soil solution of the meadow could have induced the decrease of archaeal amoA genes. Thus, our findings corroborate that responses of archaeal amoA abundances to drought could be driven by changes in soil NH4+ concentrations as a consequence of decreasing water content rather than by decreasing soil water content alone (Thion and Prosser, 2014; Delgado-Bazquerizo et al., 2013).

Overall, these findings partly support the hypothesis that the impact of drought on the abandoned site could be stronger at the meadow than at the abandoned site. In contrast to the managed meadow, where aboveground plant biomass is annually removed and harvested, at the abandoned site a thicker litter layer (Meyer et al., 2012), and higher SOM content were accumulated, which might have protected the soil from drying out (Brady and Weil, 2002; Franzluebbers, 2002; Knapp et al., 2008).

Rewetting after drought has been shown to induce short-term increases of N mineralization in soil, which are likely caused by microbial release of accumulated osmolytes to avoid lysis, but also by re-connecting soil pores and increasing the nutrient availability for microbes (Evans and Wallenstein, 2012; Fierer and Schimel, 2002; Saetre and Stark, 2005). We therefore expected a strong increase of N mineralization rates in the soils after rewetting, but at both sites one day after rewetting neither gross N mineralization, nor NH4+ immobilization rates responded to water addition, which is similar to findings by Chen et al. (2011). Due to the described short-term character of N mineralization peaks after rewetting we might have missed a possible increase.

The potential rates of nitrification, as well as of NO3- immobilization were increased at both sites one day after rewetting, while neither AOB, nor AOA gene copy numbers were affected. Placella and Firestone (2013) detected strong increases in the transcript abundances of both bacterial and archaeal amoA within two hours after rewetting at constant gene copy numbers, accompanied by increases in nitrification rates, which suggests an activity pulse of ammonia-oxidizing organisms (Fierer and Schimel, 2002). In non-drought-adapted grassland soils, however, AOB were found to have benefited from the flush of NH4+ after rewetting and to be more tolerant and resilient after drought than AOA (Thion and Prosser, 2014). Nonetheless, rewetting after long dry periods may initially rather stimulate microbial activity, and DNA-based methods could underestimate potential dynamics (Barnard et al., 2013).

In spite of the fact that effects of drought were very diverse and divergent between sites, 7 weeks after termination of drought all determined parameters, (except NO3- concentrations at the meadow), were similar to undisturbed controls, which indicates a high resilience of the studied grasslands after drought. Moreover, we detected a strong temporal variability of organic and inorganic N concentrations in soils at both sites that might be related to overall seasonal microclimatic fluctuations, and indirectly to altered levels of plant N demand and plant phenology (see e.g. Kaiser et al., 2011).