The six selected surface soils (0–20 cm) were sampled from three winegrowing areas in France and one in Italy, presenting different soil types, physico-chemical properties, and in particular different Cu treatment histories (Table 1). Among the three soils coming from the “Bordeaux” area, two of them (HBN, CO) have been used for viticulture and received Cu fungicide application for over 100 years. Replantation of grapevine was reported to be problematic for these soils, likely due to Cu toxicity to young plants (details are published in Anatole-Monier in 2014). The third soil from the “Bordeaux” area (OB) was a forest soil that was recently converted to conventional viticulture including the use of Cu-based fungicide (conversion 4 years before sampling). The two soils from the “Soave” region (CI, VI) in northern Italy have received Cu fungicide applications for about a century but there was no evidence of toxicity for grapevines. The last selected soil (STM) is a vineyard soil from the Saint Mont region (STM), which did not receive any Cu treatment.

The soil samples were air-dried and sieved to 2 mm. The soil pH was measured on 1 g in 5 mL ultrapure water (18.2 MΩ) following the ISO 11464 protocol. The cation exchange capacity (CEC) was determined using cobalt hexamine. Therefore, 1 g of soil was shaken in 20 mL of a 0.017 mol L^–1 cobalthexamine solution for 1 h, solutions were subsequently centrifuged and the supernatant filtered at 0.22 μm. Then, the cobalthexamine loss from the solution was determined by absorbance loss at 475 nm with a Varian Cary 50 spectrophotometer, Varian, Palo Alto, CA, United States. The total soil organic carbon (SOC) and total soil inorganic carbon (SIC) were calculated after subsequent measures of raw and calcined samples on a EMIA – 320 V CS automate, by Horiba Kyoto, Japan.

Composite samples (100 kg) of the first 20 cm of each soil were taken during winter and spring 2016 before the first annual fungicide treatment. The soils were root-picked, dried at room temperature, and sieved with a 2-mm mesh before potting. Soil samples were then filled into 5 L PVC (polyvinyl chloride) pots and one young vine plant (Vitis vinifera L. cv. Tannat grafted on rootstock V. riparia × V. rupestris cv. 101.14) was planted into each pot. Five replication pots were prepared for each soil modality and a microporous cup (RHIZON^® MOM 10 cm, Rhizosphere Research Products, pore size ∼0.1 μm) was inserted into each pot for sampling soil solution (Figure 2). After planting, pots were saturated with water and placed on heated ground until bud break (5d). Then, pots were placed in a greenhouse at the ISPA laboratory (INRAE Institute, Bordeaux, France) under artificial lighting mimicking 12 h of daily sunshine (Figure 2). The plants were regularly watered using demineralized water to maintain soil moisture at 80% of the water holding capacity of each soil.

Soil solution sampling began one week after potting and from then on samples were taken every two weeks. Samples were taken by applying under pressure to the suction cups for 10 min, thereby extracting between 0.5 and 10 mL of solution. The first two samples (week 0 and week 2) were not included in the analysis, to let the system equilibrate. From week 4 to 16, the soil solution samples were collected and the total organic carbon (TOC) and total inorganic carbon (TIC), as well as pH, were measured promptly after sampling. Leftover solutions were acidified with ultrapure nitric acid to 2% (v/v) for later analyses of Cu concentration and isotope ratio.

Initially, five plants were grown per modality to be sure to have at least three healthy plants, as some soils were reported to be problematic for replantation. However, only three healthy plants per soil modality were destructively harvested after 16 weeks (see Figure 2 for plants just before harvest). Five healthy fully unfolded leaves of every branch counting from the top were cut using a ceramic knife and put into plastic sample bags for Cu content and isotope ratio measurements.

Roots were extracted from soil and washed under flowing demineralized water. When no more soil particles were visible, the roots were cut from the trunk and put into sample bags. Root and leaf samples were then washed three times using demineralized water and twice using ultrapure water (18.2 MΩ). All samples were frozen at −80°C and freeze-dried. Once dried, root and leaf samples were ground to powder using a planetary mill with Zr-containers and balls.

Sample digestion was carried out in the ISO 7 cleanroom laboratories of the GET Toulouse and the LEGOS Toulouse. A 100 mg sample of each ground soil was digested in a MARS 5 microwave oven (by CEM, Matthews, NC, United States) using ultrapure acids (9 mL HNO₃: 2 mL HCl: 3 mL HF), the solution was then evaporated and the samples were dissolved in double subboiled HNO₃ for analysis. Details of the digestion protocol were published elsewhere (Blotevogel et al., 2018). The plant samples were digested on hotplates in three steps. For each plant sample, 200 mg of powder were weighed into Savillex Teflon vessels. Then, 1 mL of ultrapure hydrogen (H₂O₂) was added and left to react for 2 h at room temperature. Subsequently, 5 mL of double subboiled nitric acid (HNO₃₎ were added in 1 mL steps to each vessel and left to react overnight. Successively, the vessels were heated to 120°C for at least 4 h and evaporated to dryness at 90°C. Once dried, 4 mL of double subboiled HCl and 2 mL of double subboiled HNO₃ were added along with 1 mL of suprapure HF. Finally, vessels were heated to 120°C for at least 4 h and evaporated to dryness at 90°C. A final digestion step was performed using 5 mL HNO₃, the samples were again heated to 120°C for at least 4 h and evaporated to dryness at 90°C.

All soil solution samples with sufficient volume (more than 100 μL) were analyzed for their Cu concentration. During the whole experiment, 1 replicate for OB, 2 for CO, 5 for CI and VI, 6 for HBN, and 9 replicates for STM could not be analyzed for their Cu content, either due to handling errors or too low sample volume (see Supplementary Information for details).

In soil soultions and digested soil and plant samples, total Cu concentrations were measured by inductively coupled plasmamass-spectrometry (ICP-MS) (7500ce, Agilent Technologies – Santa Clara, CA, United States) and inductively coupled plasma – optical emission spectrometry (ICP-OES) (Ultima Expert, Horiba Jobin Yvon, Kyoto, Japan) at the GET laboratory. The analytical accuracy and precision of measurements were ensured by measuring replicates of the SLRS-5 river water standard (n = 6). Relative standard deviations (RSD%) were <5% and Cu-recovery 100 ± 5%. The accuracy and precision of the whole sample preparation procedure were checked by determining Cu concentration in BCR-2 basalt standard [Cu recovery of 96 ± 10% (n = 3)] and SRM-1515 apple leaf standard [Cu recovery of 92 ± 5% (n = 4)].

Cu isotope measurement by multicollector-inductively coupled plasma mass spectrometer (MC-ICP-MS) (Neptune, by Thermo Fisher, Bremen, Germany) required at least 500 ng of Cu. When the amount of Cu was insufficient in a single soil solution sample, we chose to pool all replicates from a given sample week (e.g., STM and VI modalities and the late samples of OB). This allowed us to have at least one soil solution Cu isotope measurement per modality at each sample time.

Purification of digested samples was carried out under ISO1 laminar flow hoods using anionic AG MP-1 resin (Bio-Rad PolyPrep chromatographic columns – Hercules, CA, United States) (Maréchal et al., 1999). Volumes of resin and solution were adapted for the different matrices to ensure quantitative elution, as the separation procedure fractionates Cu isotopes up to ≈ 19‰, between the first and the last mL of Cu elution (Maréchal and Albarède, 2002). Soil samples were purified according to the protocol described in Blotevogel et al. (2018) and displayed in the supporting information (Supplementary Information Table 1). Soil solution aliquots were digested in three steps as described above for plant samples. For purification of soil solution, root, and leaf samples, the same Bio-Rad Poly Prep chromatographic columns were filled with 2 mL of AG MP-1 resin and conditioned according to Maréchal et al. (1999). After sample loading using 1 mL of a 7 M HCl solution containing 0.01% of H₂O₂, the matrix was eluted by passing 9 mL of the same 7M HCl, 0.01% H₂O₂ solution. Cu was then eluted using 24 mL of the same solution (Supplementary Information Table 1). This same purification procedure was carried out twice for soil solution, root, and leaf samples. Details on the method development and elution profiles can be found in Blotevogel (2017). Cu isotope ratios are expressed in ‰ relative to NIST 976 Cu standard (Eq. 1).

Recovery of all purified samples was checked to be 100 ± 5%. BCR-2 standards were used as precision and accuracy control of the whole procedure, the reference material was digested three times and four different purification runs were performed allowing 10 measurements. BCR-2 isotope ratios (0.26 ± 0.09‰) were slightly heavier than values from literature 0.20 ± 0.10‰ (Babcsányi et al., 2014) and 0.22 ± 0.05‰ (Bigalke et al., 2010a) but within the 2SD range of those results.

All data analysis was carried out using the R software^® in version 4.0.3. Plant biomass and Cu concentration data were analyzed using a one-way ANOVA. Except for Cu concentrations in roots ([Cu]_Roots) and in soil solutions ([Cu]_Solution) there was no departure from normality of residues within a 95% confidence interval using a Shapiro-Wilk test. To achieve normality, the [Cu]_Roots and [Cu]_Solution datasets were submitted to a logarithmic transformation and afterward satisfied the above-mentioned criteria. The one way ANOVAs were computed to determine if the soil type had a significant influence on the data set. Differences between means were then tested using Tukey’s HSD. Differences between groups are shown in compact letter display computed using Tukey’s HSD with p < 0.05. For soil solution samples, a one-way ANOVA was computed for each soil type to detect if there were significant differences in Cu concentration over time.

Isotope data were reported using the mean value of the three replicate samples and a 2 standard deviation (SD) error bar around this mean. Differences were considered as significant if 2SD intervals did not overlap. For isotope fractionation data the standard deviation of replicate measurements was propagated.

Copper (Cu) concentrations in bulk soils ([Cu]_Soil) were between 3 and 251 mg kg^–1 (Table 2). OB and STM had the lowest [Cu]_Soil (3 and 10 mg kg^–1, respectively), whereas CO, CI, VI, and HBN soils had higher [Cu]_Soil of 115, 214, 229, and 251 mg kg^–1, respectively.

During the experiment, for each soil modality, the [Cu]_Solution remained in the same order of magnitude (Figure 3). Only for CI the one-way ANOVA showed a significant influence of the time on [Cu]_Solution (p < 0.0002) and Tukey’s HSD indicated that the solutions of week 16 had significantly (p < 0.02) higher Cu concentrations than solutions from all other time steps (Figure 3).

The lowest mean [Cu]_Solution was measured in STM soils (19 μg L^–1), followed by CI and VI soils with mean [Cu]_Solution of 40 and 100 μg L^–1, respectively (Figure 4). The highest [Cu]_Solution values were measured in soils from the “Bordeaux” area with 297, 999, and 2,705 μg L^–1 (for OB, HBN, and CO, respectively). The differences between different modalities were much larger than variations over time within the same soil (Figure 4). A one-way ANOVA showed a significant influence of soil type on [Cu]_Solution (p < 10^–15) and Tukey’s HSD indicated that mean [Cu]_Solution for each soil modality was significantly (p < 10^–5) different from all other soil modalities (Figure 4).

There was no direct correlation between [Cu]_Soil and [Cu]_Solution (Figure 5A), also between the TOC of the soil solution and [Cu]_Solution no direct correlation could be established, even when STM was excluded from the fit. Nevertheless, there was a positive relationship between mobile SOM (expressed as TOC/SOC) and mobile Cu (expressed [Cu]_Solution/[Cu]_Soil), when STM was excluded (Figure 5B, R² = 0.92, p = 0.01, n = 5).

Mean Cu isotope ratios in soil solutions include all measured samples with the exception of week 4 (Table 2). Samples of week 4 were excluded because they appeared to show transitional variations in some samples as discussed below. For STM, VI, and late OB samples all replicates were pooled for a given time step. To detect temporal variations, the evolution of one replicate is shown for CI, HBN, and CO. In STM, VI, and CI soil solutions, Cu isotope ratios are heavier than in the corresponding bulk soils (Δ⁶⁵Cu_{Solution–Soil} around + 0.4‰, Figure 6). No Cu isotope fractionation between soil and soil solution was detected in soils from the Bordeaux area (Figure 6). Only in the first analyzed samples (week 4) of HBN and VI, δ⁶⁵Cu values were significantly heavier in soil solution than in bulk soil (Δ⁶⁵Cu_{Solution–Soil} of 0.29 ± 0.09‰ and 0.63 ± 0.03‰, respectively).

After 16 weeks of growth, the measured root biomass was between 10.1 and 20.9 g DW (Figure 7). A one-way ANOVA showed a significant influence of soil modality on root biomass (p < 0.02). Tukey’s HSD showed that only the mean root biomasses of VI and OB soil were significantly different (p < 0.04), with 6.28 g higher root biomass in VI than in OB. For all other samples, differences were not significant (Figure 7).

The leaf biomass was also different according to the soil modality (one-way ANOVA p < 0.001). Tukey’s HSD showed that CI samples had a significantly higher leaf biomass than STM (−11.3 g DW, p < 0.04), OB (−12.7 g DW, p < 0.02), and CO (−18.6 g DW, p < 0.001) samples, and VI samples had a higher leaf biomass than OB (−11.0 g DW, p < 0.05) and CO (−16.9 g DW, p < 0.03) samples. Furthermore, HBN samples had higher leaf biomass than CO (−12.2 g DW, p < 0.03) samples. Significantly different groups are shown in compact letter display in Figure 7.

The lowest mean Cu concentrations in roots ([Cu]_Roots) were detected in STM and OB grown plants with 30 and 25 mg kg^–1 (Table 2), followed by the Italian soils CI and VI, with [Cu]_Roots of 199 and 81 mg kg^–1. The highest [Cu]_Roots were found in CO and HBN soils with, respectively, 579 and 768 mg kg^–1. The differences in [Cu]_Roots were statistically significant between the soil modalities (one-way ANOVA p < 10^–9). Tukey’s HSD showed that with exception of the pairs OB-STM and HBN-CO, [Cu]_Roots significantly differed between each soil modality (p < 0.002, Table 2). The [Cu]_Roots followed the rank of [Cu]_Solution (Figure 4) for most soils, only the rank of OB decreased. Except in CI and VI, [Cu]_Roots was higher than the [Cu]_Soil so that significant contamination by possible leftover soil particles can be excluded. [Cu]_Roots was not correlated to [Cu]_Soil or [Cu]_Solution (Figures 8A,B).

Copper isotope fractionation between soil solution and roots was increasingly positive with higher [Cu]_Solution (Figure 9A). Negative Cu isotope fractionation between roots and soil solutions occurs in STM, VI and CI modalities (Δ⁶⁵Cu_{Roots–Solution} of -0.36 ± 0.09‰, -0.26 ± 0.12‰, and -0.18 ± 0.11‰, respectively). Positive isotope fractionation between roots and soil solution was observed for the Bordeaux soils OB, HBN, and OB (Δ⁶⁵Cu_{Roots–Solution} = 0.05 ± 0.16‰, 0.25 ± 0.23‰ and 0.26 ± 0.21‰, respectively). A logarithmic function (Δ⁶⁵Cu_{Roots–Solution} = 0.137 * ln ([Cu]_Solution) –0.766) was fitted through all data points (n = 6, R² = 0.96, p < 0.001) (Figure 9A).

Given the large differences of Cu content in soil solutions, between 19 and 2,705 μg L^–1, the [Cu]_Leaves were surprisingly similar in the different modalities, between 4.9 and 10.9 mg kg^–1. One-way ANOVA showed nevertheless a significant influence of soil modality on [Cu]_Leaves (p < 0.04). Tukey’s HSD indicated that [Cu]_Leaves only differed significantly between STM and CO samples (p < 0.04).

When Cu content in roots was low (i.e., OB and STM), no Cu isotope fractionation between roots and leaves occurred (Figure 9B). Modalities with higher Cu content in roots (CI, VI, HBN, and CO) showed light Cu isotope enrichment in leaves compared to roots (Figure 9B).

[Cu]_Soil of OB and STM was below average European background concentration of about 14 mg kg^–1, whereas CO, CI, VI, and HBN soils showed Cu concentrations above the European background level, similar to other Cu contaminated vineyard soils (Chaignon et al., 2003; Lado et al., 2008; Reimann et al., 2018). CI, VI, and HBN exceeded the predicted no-effect concentration (PNEC) for Cu in soils (200 mg kg^–1) but concentrations were still two to three times less than maximum Cu contents measured in European vineyard soils (Ruyters et al., 2013; Reimann et al., 2018).

Copper mobility in soil solution appeared to be controlled by SOC mobility, rather than by the amount of TOC in the soil solution (Figure 5B). It has long been established that Cu in soil solution is primarily bound to organic matter, however, the role of the mobility of SOC on Cu mobility was less studied (Yin et al., 2002; Chaignon et al., 2003; Ashworth and Alloway, 2004; Ren et al., 2015). In the present study, only soil solutions from the STM soil did not fit into the trend (Figure 5B). This might be linked to the lower pH of STM soil solution, and thus slight differences in Cu speciation in solution or competition between Cu and protons for adsorption sites on dissolved organic matter (Yin et al., 2002; Chaignon et al., 2003; Bravin et al., 2009a,2012). STM is also the only soil that did not receive any Cu treatment so that it is possible that only exogenic Cu is controlled by SOM mobility (Wang and Staunton, 2006). This is supported by a correlation between Fe and Cu contents in the STM soil solution (Supplementary Information).

A mobile Cu pool, adsorbed or complexed with organic matter is consistent with a heavy isotopic Cu signature in solution, as observed for STM, CI, and VI (Figure 6; Yin et al., 2002; Chaignon et al., 2003; Ashworth and Alloway, 2004; Vance et al., 2008; Bigalke et al., 2010a; Kusonwiriyawong et al., 2016). The absence of isotope fractionation between soil solution and bulk soils from the Bordeaux area suggests that the total Cu pool in soil and the mobile Cu pool have similar chemical speciation or that desorption does not fractionate Cu isotopes (Figure 6). No isotope fractionation between soil and soil solution would for example be observed if Cu gets into the soil solution by solubilization of the organic molecule to which Cu is bound.

The absence of a [Cu]_Soil – [Cu]_Solution correlation (Figure 5A) is in contrast with previous studies that found correlations between bulk soil Cu, extractable Cu, or the ionic Cu²⁺ fraction and root Cu (Chaignon et al., 2003; Bravin et al., 2009a,b, 2012). Some studies reported a good correlation between bulk soil Cu and pore water Cu, especially in contaminated soils (Sauvé et al., 2000). However, our observation is perfectly in line with research showing that pH and surface adsorption control Cu mobility, not bulk soil content (Brun et al., 2001; Yin et al., 2002; Chaignon et al., 2003). There was little variation in [Cu]_Solution over time compared to water-extractable Cu in soil incubation experiments (Wang and Staunton, 2006). The increase of [Cu]_Solution in CI over time might reflect active processes, including rhizosphere acidification, exudation of reducing agents or phytosiderophores, implemented by plants to satisfy needs of Fe nutrition (Kraemer, 2004; Ström et al., 2005). Indeed, CI is the most calcic soil of the experiment and Fe deficiency is common in such soils (Kraemer, 2004; Ström et al., 2005).

In our setting, the offset in isotope fractionation in the first sample is likely due to the rewetting and adaptation of the soil to the new environment, accompanied by a priming effect (Figure 6; Miller et al., 2005). It was formerly reported that the initial DOM released from rewetted soils has a lower affinity to Cu compared to later, more humified DOM (Amery et al., 2007). Detailed interpretation of this effect without further data would be speculative but if present, it suggests that transitory effects on microbiological activity and/or soil organic matter quality can play an important role in Cu mobility and isotope fractionation.