Responses of soil nitrification activities to copper after a moisture stress

Some steps of the soil nitrogen (N) cycle are sensitive to environmental pressures like soil moisture or contamination, which are expected to evolve during the next decades. Individual stresses have been well studied, but their combination is not yet documented. In this work, we aimed at assessing the importance of the soil moisture on the impact of copper (Cu) contaminations on the N cycling soil function using the potential nitrification activities (PNA) as bioindicator. A two-step experiment was performed. First, a loamy soil was incubated 5 weeks in either 30, 60, or 90% of its water holding capacity (WHC) or alternating drought and rewetting periods. Thereafter, soil samples were exposed to a gradient of Cu concentrations through a bioassay involving nitrification. The dose–response curves of PNA in function of added Cu were modeled to calculate the effective Cu concentrations, namely ECx with x being the percentage of PNA inhibition. These values were then compared between experimental conditions to highlight differences in threshold values. The preincubation moisture treatments significantly affected the PNA responses to the secondary Cu stress with, for instance, hormetic responses in all cases except for the dry-rewetting treatment. Small PNA inhibitions were estimated for high Cu doses in the soils with low water contents (30% WHC) or submitted to dry-rewetting cycles, contrarily to the patterns observed for the soils with high water contents (90% WHC) or submitted to a single period of drought. Overall, significant differences were found in estimated ECx values between moisture treatments.


Introduction
With the current scenarios of climate change, rainfall patterns are expected to change during the next decades (Lee et al. 2014) with more intense and longer drought periods followed by intense rainfalls. These modifications in the rainfall patterns may impact the soil water contents during critic periods. Excess of water or drought may affect soil moisture which is one of the main drivers of the soil microbial activity (Moyano et al. 2013). In parallel, human activities have dispersed significant quantities of contaminants into the environment, such as trace elements which are persistent and potentially toxic for the life soil biota (Giller et al. 2009). Nowadays, the contamination of soils by trace elements coming from atmospheric source or through agricultural practices has become a major concern at a global scale (Song et al. 2012). Trace element contaminations affect several environmental processes such as those performed by soil microorganisms (Giller et al. 2009), and soil microbiological indicators (abundance, diversity, or activity) are hugely used to assess soil contamination. Microbial activities related to specific narrow niche processes are among the most sensitive endpoints to assess soil contaminant impacts (Broos et al. 2005). Therefore, bioassays measuring activities performed by specifics microorganisms are useful to assess the severity of stress encountered by the soil ecosystem and possible outcomes on soil functions.
Global climate change and diffuse contamination both affect large part of the Earth. Soil microbial communities may independently encounter single stress associated to global change or contamination, but also multi-stress due to their combined effects. Some information are known on the single-stress effects (Bååth 1989;Schimel 2018), but the multistress effect due to their interaction is less documented. Nitrogen (N) cycle is known to be highly sensitive to soil moisture (Stark and Firestone 1995). Indeed, processes of nitrification (oxidation of ammonia to nitrates with nitrous oxide production) and denitrification (reduction of nitrate to nitrogen and consuming nitrous oxide) are performed by communities that concur in soils depending on their oxygenation and thus of their moisture (Van Groenigen et al. 2015). Soils nitrifying communities have been shown to be highly sensitive to soil moisture variations (Guo et al. 2014;Van Groenigen et al. 2015) but also to the presence of heavy metal in soils (Ruyters et al. 2010;Smolders et al. 2001). Moreover, N cycle is performed by a limited number of microorganisms, making it more sensitive to environmental pressures than carbon cycle (Broos et al. 2005). In the literature, this soil function concerning N cycle is often assessed from the measure of the potential nitrification activity (PNA) (Broos et al. 2005).
However, effects of a primary stress on the response of microbial functions to a secondary stress are not well known. Depending on the stress and their order of appearance, different type of microbial responses were reported in the literature (Philippot et al. 2008;Rusk et al. 2004). It is thus difficult to predict the outcomes of combined effects such as ecological stress due to climate change and chemical stress due to the presence of contaminants in soils.
In this context, this work aims at investigating relationships between moisture stress as example of climate change stress and soil contamination pressure on soil functions performed by soil organisms. For that, we chose to assess the primary effect of the soil moisture history followed by a copper (Cu) stress on the soil PNA. Cu was used to mimic soil contamination, and PNA was used as a proxy of the soil nitrification process. We focused on the effect of newborn exogenous Cu contaminations brought through a bioassay to a soil that was first submitted to various moisture stresses. Indeed, taking into account that soil nitrifying and denitrifying organisms compete for oxygen, we assumed that soil pre-exposition to various moisture promoted one process regarding to the other and then the response to the secondary stress (Gleeson et al. 2008;Li et al. 2016). We hypothesized that less favourable pre-incubation periods for organisms' activities in terms of moisture affect the microbial community function related to the N cycle and PNA values with a decrease in effective Cu concentration threshold values when a stress on stress occurs. For that, a dose-response approach was carried out. First, absolute PNA (aPNA) responses after moisture then Cu stresses were quantitatively analysed. Then, relative PNA (rPNA) responses compared to the noadded Cu sample were used to both quantify the effective Cu concentrations inducing 5, 10, 20, and 50% (EC 5,10,20,50) of decrease in rPNA and discuss the pattern of the dose-response curves.

Soil sampling
The soil, described in Obriot et al. (2016) as a luvisol with 11 g/kg of organic carbon and 1.0 g/kg organic nitrogen (determined according to NF ISO 10694 and NF ISO 13878) and pH 6.9 (pH in water, determined according to NF ISO 10390), was sampled in January 2017 in the control plot of the experimental "Qualiagro" site (48° 87′ N, 1° 97′ E 17). This agricultural plot is not contaminated with Cu. Its Cu content around 12 mgCu.kg −1 is consistent with the regional pedogeochemical background, and the fact that no inputs were registered since more than 20 years. Several fresh soil samples were pooled and immediately sieved at 5 mm and stored at 4 °C few days before building-up the microcosms to drive soil moisture regimes. Aliquots of this sieved soil were used to measure the fresh soil moisture at the time of sampling using three replicates of 5 g of fresh sieved soil dried at 105 °C during 48 h. Soil humidity at the time of sampling was thus 15%. Later on, all values are expressed on the dry weight basis. The field capacity as maximum water holding capacity (WHC) was defined by the total amount of water the soil can hold without losing it by drainage and was set up at 32% (w/w) on a gravimetric basis.

Experimental setup
To evaluate the impact of soil moisture on the sensitivity of nitrification to Cu toxicity, a two-step experimental approach was carried out. A primary pre-incubation period of 5 weeks was running under different moisture stresses, and a secondary metal contamination stress was applied, through a bioassay that lasted 72 h and allowed evaluation of the absolute and relative potential nitrifying activities in the presence of a copper gradient.

Microcosms to drive soil moisture regime
Soil microcosm consisted of about 500 g of soil in plastic boxes maintained half-open during 35 days and submitted to different treatments. Starting from the known initial moisture, three microcosms were set up at 30%, 60%, or 90% of the soil WHC, respectively, in order to roughly span respectively limiting, optimal, and close to water saturation treatments for the microbial activity. These WHC were kept constant by weighting all along the incubation. Later on, these three treatments will be called "30%, 60%, and 90%," respectively. Two others microcosms were incubated with variable water content in order to simulate two kinds of drought and dry-rewetting cycles. One was left for about 3 weeks dry period without water inputs until 10% of the WHC before progressively rewetting at 60% WHC, while the other was treated with two dry and wet cycles alternating one-week dry period (10% of the WHC by air-drying) followed by 1 week near-saturation period (90% WHC). Drying was performed by natural evaporation (gentle air-drying at the laboratory temperature, i.e., 20 °C) and moisture control was controlled by weighting. The two last samples will be called thereafter "Drought" (DO) and "Dry-rewetting" (DR). The concentrations of N-NO 2 , N-NO 3 , N-NH 4 , and dissolved organic carbon (DOC) were measured at the end of the pre-incubation period for the controls without added Cu and for each of the five moisture treatments. The ammonium ion NH 4 + was extracted on the basis of a ratio soil:solution of 1:10 with KCL 0.5 M and the N-NH4 + content was determined a 610 nm with an HACH reagent (Loveland, CO, United States) after centrifugation. Nitrate (N0 3 − ) and nitrite (NO 2 − ) ions were determined by colorimetry according to the Griess reaction (Miranda et al. 2001). The supernatants were dropped in microplates and the Griess solution was added (HCl, 0.5 M, vanadium chloride III (Sigma-Aldrich 208,272) at 1 g L −1 , sulfanilamide (Sigma-Aldrich S9251) at 2.5 g L −1 and N-(1-naphthyl)-ethylenediaminedihydrochloride (Sigma-Aldrich 222,488) at 0.25 g L −1 ) and then incubated at 60 °C for 1.5 h. The optical densities were determined at 540 nm with a microplate reader (SAFAS Xenius, Monaco). The DOC concentration was determined using a TOC analyser (Schimadzu, TOC-L). Results are presented in Table 1.

Bioassay to assess further Cu impact
After the pre-incubation period, the soil bioassays were immediately performed by measuring nitrate (NO 3 − ) production rates over a short-term aerobic incubation in soil slurries (ratio soil to solution, 1:10) with ammonium in excess and in the presence of gradients of Cu concentrations. For exposure, copper was used as a CuCl 2 -salt (Sigma-Aldrich, Purity > 99.5%). The soil potential nitrification was then calculated on the basis of NO 3 − measurements over the time period. The PNA bioassays were adapted from the methods proposed by (Petersen et al. 2012). Briefly, 3.5 g of fresh soil (approximately 3 g of soil equivalent dry weight) was mixed in a 50-mL Falcon® tubes with 29 mL of a 10-mM HEPES buffer solution (hydroxyethyl piperazineethanesulfonic acid, Sigma-Aldrich, France) to maintain a constant pH under Cu spiking and nitrification activity, and containing the substrate (NH 4 ) 2 SO 4 (3 mM) (Sigma-Aldrich, France). Then, 1 mL of Cu solutions at different concentrations were added to reach final added Cu concentrations of 50, 100, 250, 500, 750, 1000, and 2000 mgCu. kg −1 soil. Controls with only Milli-Q® water (Millipore) were also performed. Three independent samples (n = 3) from each pre-incubation treatment were ran in three concomitant bioassays for each Cu concentrations. Following an initial horizontal shaking step (250 rpm, 10 min), the soil slurries were left incubated on a rotary shaker (150 rpm), under aerobic treatments, at 25 °C during 72 h. After 10 min, 24 h, and 72 h of incubation, 2-mL aliquots of the soil-solution mixture were transferred in Eppendorf® vials and centrifuged for 5 min at 13,000 g at 4 °C. The supernatants were collected and stored in microplates at -20 °C until analyses of NO 3 − and NO 2 − by colorimetric determinations, following the reduction of NO 3 − in NO 2 − by vanadium(III) and then the detection of NO 2 − by the acidic Griess reaction (Miranda et al. 2001). Three different aliquots from each bioassay tube were analysed. Finally, the absolute value of PNA, aPNA in µg N-NO 3 g −1 soil h −1 , was calculated on the basis of N-NO 3 − + N-NO 2 − concentrations measured at the different time steps. The points Cu = 0 allowed us to verify that NO 2 − contents were negligible (Table 1), so that

Statistical analysis for the different moisture treatment on Cu comparison
Data of aPNA values between the different moisture treatments were compared for each Cu level with the use of Kruskal Wallis test followed by a post hoc Dunnett test. Adjustment of p-value for multiple comparison was made by Holm procedure to limit false negatives.
Dose-response curves for rPNA were analyzed with the DRC package, following recommendations of (Ritz 2010;Ritz et al. 2015). For the dose-response curves fitted by the same functions, the parameters were compared using the compParm function of the DRC package (see paragraph Dose-response curve analysis).
ECx comparisons were made through the estimated confidence intervals (IC), so that for each ECx, the moisture treatment with overlapping IC was not significant different. All analyses were conducted using R v4.1 (R Core Team 2021).

Dose-response curve analysis
To fit the dose-response curves, we have proceeded by different steps according to our aims of both estimate and compare the variations in rPNA due to Cu contamination stress under the different moisture stresses. Expected results are to provide EC5, EC10, EC20, and EC50 values as % response (at 95% confidence interval) expressed as rPNA for each moisture treatment.
For the first aim, i.e., estimating the variations in rPNA, the selection of the best dose-response model able to fit the experimental data was achieved by testing the following widely used models in ecotoxicology studies (Ritz et al. 2015): the log-normal (LN) dose-response model, or the log-logistic (LL) and their Weibull derivative (W1 or W2), or the hormetic models as the Cedergreen models (CD, tested with alpha = 1 and alpha = 0.25), the Gompertz (G) model, or the Braincousens models (BC.4 or BC.5 depending of the number of parameters to be fitted). Lower Tfinal − Tinitial × VsW asymptotes were fixed to 0 and upper to 1 except for the hormetic models. For the second aim, we considered that the comparison of the dose response curves and their parameters as a whole between all the moisture treatments better required using a common model for all the curves, contrarily to ECx determinations that required more precise fits to extract ECx values. We thus compared the five best models per moisture treatment according to AIC and logLik criteria and confirmed the choices by visual confidence interval accordance (limited confidence interval). From these five pre-treatments × 5 models, we looked for a common model. If no common type of model was found powerful enough to extract ECx values, we selected more types of models for these extractions.

Results
Evolution of the absolute PNA values under successive stresses are given in Fig. 1, with the mean, 1st and 3rd quartile obtained for all pre-incubation treatments (DR, DO and, 30, 60 and 90% WHC treatments) in function of the Cu concentration gradient.
The comparison between the control bioassay without Cu (Fig. 1, Cu 0) and the values in the presence of added Cu shows that the moisture pre-incubation treatments significantly affect the aPNA. Significant differences were found between aPNA for soils incubated under DR stress and soils incubated under 60% WHC, with aPNA values roughly smaller by one third in the soils submitted to dry and rewetting cycles (DR treatment).
The aPNA values decreased with increasing soil Cu concentrations and the moisture pre-treatment differently affected the aPNA inhibition by Cu stress. At the highest Cu doses (> 1000 mgCu.kg −1 ), no more differences in aPNA values could be observed whatever the moisture treatments. Therefore, above a threshold of 1000 mgCu.kg −1 , the effects on the nitrification processes were the same whatever the initial moisture pre-incubation.
Evolution of the relative PNA values under successive stresses was analyzed by fitting the dose responses curves. As a first step, we aimed at selecting models to fit the dose-response curves for ECx estimation. Suppl. table 1 gives the values of the criteria obtained for the five best fit of each of the dose-response curves: rPNA inhibition = f (total added soil Cu). Model selection based on the lowest value of AIC and higher value of LogLikelihood criteria showed that Cedergreen (alpha = 1) and Braincousens or Weibull models are the best ones describing the inhibition of rPNA in function of total added Cu for incubation performed under constant moisture treatments (Supp Table 1a). Cedergreen and Braincousens models integrate hormesis parameters, so that modeling included our observations of an increase in rPNA at small Cu inputs in several moisture treatments (Fig. 2a).
If we consider the AIC criterion, inhibition of rPNA in DO incubated soils were better described by Braincousens or Weibull II and DR by Weibull I models (Suppl. table 1b). Fig. 1 Mean, 1st and 3rd quartile of aPNA measured at the end of the bioassay for each incubation treatment and added Cu concentration. Preincubation treatment is represented per color with 30%WHC in green, 60% in orange, 90% in purple, DR in yellow, and DO in brown. For each Cu level, post hoc Dunett test with Holm procedure was applied. Significant differences in PNA between moistures treatments are indicated by brackets branches and notated with stars (*p.v < 0.05) Fig. 2 Fit of the selected function on the AIC criteria to model rPNA responses to added Cu with 95% confidence interval. a In the case of the incubation performed at constant moisture (30% in green, 60% in orange and 90% WHC in purple) with fit of Cedergreen functions. b In the case of the incubation performed under DR (dry rewet, yel-low) and DO (dry only, brown) moisture treatment with fit of Weibull functions. Weibull model of the first type was used in the case of the dry-rewet incubation, and Weibull model of the second type was used in the case of the dry only treatment But for DO the Weibull fit indicators (AIC, Lack of Fit…) were close to those of Braincousens. We chose using Weibull for homogeneity with the DR modelling to further extract ECx values. This means that rPNA inhibition of soils incubated under DO treatment was poorly affected by the Cu gradient for low Cu concentrations. On the contrary, soils incubated under DR treatment are largely affected by variations of moderate Cu concentrations but poorly by the highest (Figs. 1 and 2b).
In order to compare and assess the effect of pre-incubation on the dose-response curves of rPNA against total added Cu concentration, all the experimental curves had to be fitted with a single model that takes into account biological meaning. For that, we selected the Cedergreen model applied to all curves to compare the parameters of all the dose-responses models. Indeed, we tested specifically the Cedergreen modelling function in the case of DO and DR conditions for two reasons: (i) Braincousens and Cedergreen models are both hormetic models and (ii) the Cedergreen fits were found acceptable compared to Braincousens for all the constant moisture treatments.
In the case where all the experimental curves were fitted with the Cedergreen model to assess the effect of the pretreatment, the Cedergreen parameter models obtained are given in Table 2. They were found significantly different from 0 (p < 0.05) for all pre-incubation treatments except the hormesis effect in DR. This means that stimulation of the rPNA at low Cu concentration is not significant in the case of DR treatment (Table 2, Fig. 2b). No differences in the size of the hormetic effect (f parameter, Table 2) were found between all the four other treatments. For these four other treatments, we were able to estimate the maximum estimated response (e.g., the maximum increase in the potential nitrifying activities) and the associated Cu concentration. In all cases (30, 60, 90% WHC and DO), the increase in rPNA at small doses of added Cu is limited to 5% (maximal response expected from 1.047 to 1.052% of aPNA). The associated Cu concentration modeled by the Cedergreen fit was roughly twice higher for the 90% WHC and the DO treatment than for the 30 and 60% WHC treatments (22 and 26 mgCu.kg −1 against 11 and 14, respectively).
In parallel, the "b " parameter which represents the steepness of the curve after hormesis is found significantly higher in the DO case than in the four other treatments, whereas it was found significantly smallest in the 30% WHC case. This means that for DO treatments the decrease in rPNA is steeper for Cu concentration above dose of maximal response (26 mgCu.kg −1 ) than for the other condition and that for 30% WHC the decrease in rPNA is smallest for Cu concentration above dose of maximal response (11 mgCu. kg −1 ).
In a second step, we aimed at estimating the effective Cu concentrations ECx inducing x% of PNA inhibitions in each pre-treatment cases. Therefore, more precise dose-response curves are required for each pre-incubation condition. Hence, we used Cedergreen models for the constant moisture pre-treatments and Weibull models for the pre-treatments with various moistures (Weibull II for DO, Weibull I for DR). Results for EC5, EC10, EC20, and EC50 for each soil moisture status are given in Table 3.
For a given percentage of rPNA inhibition, we found significant differences in the estimated amounts of Cu inhibiting rPNA between the different moisture pre-treatments (Table 3). To observe a 5%inhibition of rPNA, EC5 predicted a low soil Cu content of 185 (± 66) mgCu.kg −1 for soils pre-incubated at 30% WHC compared to a high soil Cu content of 405 (± 88) mgCu.kg soil −1 for soils pre-incubated under DO treatment. Similarly, a 20% rPNA inhibition was predicted for a low soil Cu content of 440 (± 88) mgCu.kg −1 for the 30% WHC pre-treatment compared to a high soil Cu content of 721 (± 95) mgCu.kg soil −1 for soils pre-incubated under DO treatment. This pattern is found also for EC10 but not for the 50% level. Taken into account the lower and upper values of EC5, we estimated that rather small Cu total concentrations (around 185 mgCu.kg −1 ) added to soil are able to induce a decrease in potential nitrifying activities when the soils are pre-incubated at 30% and 60% WHC but not for soils at 90% WHC or when soil is subjected to DO pre-treatment. In the same manner, we estimated that a value Except for x = 50, the ECx values calculated for DO were always found higher than for DR. However, the differences were only significant in the case of EC5. But for a given preincubation, we observed no significant differences between the estimated EC5 and 10 (Table 3). For x = 50, the EC50 values were found significantly higher after the DR pre-treatment (1763 ± 230 mgCu.kg −1 ) compared to the 30% WHC one (1220 ± 179 mgCu.kg −1 , Table 3).

Use of PNA as indicator to derive ECx values in the case of successive moisture and chemical stress events
PNA is a frequently used indicator of soil contamination and of loss of soil functions (Broos et al. 2007; Hund-Rinke and Simon 2008) despite its known high variability in nonpolluted soils (Sauvé et al. 1999). Here, we submitted a soil sample first to a moisture stress before applying a secondary metallic stress, and we focused on the potential soil nitrification activities responses to these two conjugated stresses. Absolute PNA measured in our bioassay is a reflection of the dominant activity of the nitrifying communities selected through microcosm moisture treatments. In complement, rPNA refers to soils' ability to resist to a subsequent Cu stress.
Dry-rewetting cycles for soils have been reported to enhance organic matter mineralization inducing C and N flushes with rewetting (Birch 1958). But various effects on the resulting NO 3 − concentrations were observed, depending on the soil management or on the number of dry-rewetting cycles (Fierer and Schimel 2002). It is thus difficult to predict the nitrification process under various moisture treatments. In our study, and for one given type of soil, we compared the nitrification process at the end of different moisture pre-incubations (Table 1). Dry rewetting cycles do not seem to induce carbon flush and higher nitrification process in function of the produced NH 4 + was observed in DO and even more in DR microcosms. Overall, our results are consistent with the fact that dry-rewetting cycles play a major role in the N cycle compared to constant moisture. Our hypothesis was that the various moisture stresses do not select the same communities inducing different nitrate production abilities. Gain in knowledge would imply measures to distinguish between variations in activity or in community composition due to pre-incubation moisture stress. Nevertheless, the roughly 30% inhibition of aPNA in the soils incubated under DO or DR treatments compared to the soils incubated at a constant moisture is consistent with a primary effect of the pre-incubation condition on soil nitrifying activities. Furthermore, the variation in aPNA due to pre-incubation is on the order of magnitude than most of PNA variability observed due to soil or climatic variability (Mertens et al. 2007;Tang et al. 2019). PNA measurements can also be used as a sensitive tool to define ecotoxicological guidelines (Broos et al. 2005). In the literature, it is often noticed that EC50 values are difficult to obtain mostly due to the fact that dose-response curves are rarely fully complete until the end-point being zero (Ritz et al. 2006;Sebaugh 2011). Determination of EC50 values by extrapolation of the fit of non-complete curves could thus be less interesting and more ambiguous than the determination of threshold with smallest ECx with x < 50 using experimental points. In our experiments, we never reached a low plateau in rPNA inhibition, and we only measured two cases where rPNA was inhibited higher than 50% whatever the pre-incubation treatment. Despite the fact that satisfying at least one of these two conditions is a guarantee to provide accurate estimates (Sebaugh 2011), our results showed limited incertitude around the estimated ECx values. For the high EC values, we obtained an incertitude of 14% in concentration of added Cu in mean, compared to the 31% of total added Cu in mean for the small EC (EC5).
For the small concentrations of added Cu, the beginning of the dose-response fits showed hormesis effects with different ranges (Table 2 and estimated max PNA). An increase in nitrification activities have already been observed under various concentrations of Cu or Zn (He et al. 2018;Montoya et al. 2018). This variability in the range of the small dose effects is not captured by ECx values with small x, but rather lead to large incertitude's in EC5 determinations. Such incertitude decreased for EC10 and EC20. Thus, if ECx data are useful threshold providing values easily transferable in terms of soil contamination management, their theoretical determination after modeling may be not sufficient to identify smallest dose effects. Our results show ECx values ranging from few hundreds (184 to 404 mgCu. kg −1 for EC5) to few thousand (1221 to 1776 mgCu.kg −1 for EC50). Several studies highlighted the importance of soil type and soil properties in ECx determination (Broos et al. 2005;Rooney et al. 2006;Criel et al. 2008;García-Carmona et al. 2019). Compared to other studies using soils with pH and organic carbon contents close to ours, our EC50 values are in the higher range of those of Rooney et al. (2006) (from 187 to 359 mgCu.kg −1 for barley root elongation and 351 to 851 mgCu.kg −1 for tomato shoot biomass) or of those of Criel et al. (2008) from 250 to 500 mgCu.kg −1 for Eisenia fetida tests). But they are in the lower range of EC50 values determined by García-Carmona et al. (2019) for soil respiration (> 4000 mgCu.kg −1 ).
In our conditions, the use of PNA as indicator of successive moisture and chemical stress events was powerful to highlight an effect of moisture on the response of a soil to a Cu stress. PNA values allow comparison of ECx values between different moisture pre-treatments in particular for x = 20. However, due to the high Cu concentrations needed for experimentally reach a high PNA inhibition, no low plateau has been reached. Thus, cautious have to be taken if such values are used for policies because other soil functions might have been inhibited before the threshold is reached.

Effects of a stress on stress, as preliminary moisture followed by Cu stresses
In the present laboratory study, the Cu stress was applied secondary to events of dry-rewet or constant moisture treatments during 5 weeks. In the field, it is more likely that dry-rewet events occur in already Cu contaminated soils, which is another scenario as the one we studied. In the case of already field-contaminated soils, studies concerning the effect of soil moisture suggest low effect on metal speciation consistent with metal co-tolerance mechanisms occurring in long-term metal polluted microbial communities (Zheng and Zhang 2011;Morawska-płoskonka and Niklińska 2013). Contrarily, our results show that moisture soil history changes the soil PNA response to a supplementary metal stress. Soil samples submitted to a single long dry cycle (DO) or staying at 90% WHC seemed to be more tolerant to a subsequent Cu stress than for those submitted to dry-rewetting cycles (DR) or 30% WHC (Fig. 1, Table 3). The 60% WHC microcosms were found in-between these two cases. This can be clearly seen through the higher EC5 to EC20 values for DO soils.
The rather surprising result concerning DO could be due to our experimental design where the last week the sample was gently moistened back from 10 to 60%WHC. For DR, we hypothesized that the pre-treatment strongly selected communities. Indeed, it has been shown that less diverse microbial communities may be more sensitive to subsequent stress (Hallin et al. 2012). The hypothesis of a high primary stress in the soil incubated under DR pre-treatment is somehow supported by the absence of hormesis, whereas in the four others treatments low Cu concentrations induced slight increases in PNA suggesting that microbial communities could have enough resources to do so.
For the soil sample at 30% WHC, no effect of pre-incubation period was noticed on PNA without added Cu. But we noticed a high sensitivity to Cu contamination that could be due either to a low pool of microorganisms resistant to Cu or to a high level of Cu bioavailability as shown in suppl. Figure 1, taking into account that the Cu contents in solution mimic Cu bioavailability.
On the contrary, the soil sample incubated at 90% WHC showed high resistance to added Cu that could either be due to a lower Cu availability in the soil solution or to the presence of a higher active pool of soil bacterial communities resistant to Cu. Indeed, our Cu in solution measurements shown in suppl. Figure 1 suggest this lower availability. Unfortunately, the high variability in available Cu regarding the total Cu measurement for this condition and the lack of replicant does not allows to draw robust conclusion. To disentangle between these hypotheses, a new design of experiments would include a proxy of Cu availability as well as characterization of the communities (structure or diversity) after the first stress to assess potential differences in the modifications of microbial communities between the pre-incubations.
Finally, the aPNA values were found different between the constant and various moisture pre-treatments only for the lowest part of the dose-response curves, i.e., for low added Cu concentrations, and became similar at highest added Cu concentrations (Fig. 1). Such a result suggests that the preincubation patterns have modified the sensitivity (Cu concentration initializing a loss of function) but not the resistance (disappearance of the function) of aPNA. However, we cannot conclude if the aPNA measured under 2000 mgCu. kg −1 are minimal values (Fig. 1) due to microbial highly resistant groups or if higher levels of Cu would have reduced aPNA suggesting that even higher Cu concentration should be used to assess PNA resistance to double stress.

Conclusion
Our study showed that the pre-incubated soil samples at low moisture (30% WHC) were more sensitive to a secondary Cu stress than those pre-incubated at a higher moisture (60 and 90% WHC), with EC5 values defined respectively around 185 (± 66), 231 (± 75), and 349 (± 96) mgCu.kg −1 . Soil samples submitted to gentle drought (DO) and then gentle rewetting were surprisingly less sensitive to Cu with a EC5 value at 404 (± 88) mgCu.kg −1 whereas soil samples submitted to drastic changes in moistures (DR) lost nitrification activities as soon as low added amounts of Cu (EC5 = 189 (± 78) mgCu.kg −1 ) indicating their sensitivity to the second stress.
The extrapolation of our results obtained in controlled conditions to the field is not straightforward since Cu contamination is generally more diffuse. But these Cu amounts in agricultural parcels are not seldom with Cu inputs through fertilizers or pesticides resulting from several years of cumulative inputs (Panagos et al. 2018). Our results showed that differences in PNA values between the moisture histories we studied decrease when the Cu contamination increases. We also showed that ecotoxicological studies based on ECx determinations should be completed by dose-responses curves fitting analysis in order to highlight patterns that are more precise. Indeed, our modeling approach allowed us to emphasize small increases in PNA for low added Cu concentrations close to 20 mgCu.kg −1 in the case of four to five preincubation treatments that were not exemplified through the only ECx determinations. Thus, our results showed that climate change and particularly modifications in rainfall patterns have to be considered when dealing with N cycle. Indeed, micro-organisms and the soil functions they provide may be differentially affected by a metallic stress depending on the prevailing soil moisture history.