The Liège Province, in southeastern Belgium, was the main centre of metallurgical activities in the country. In this area, soils were largely affected by heavy-metal (Zn, Pb, Cd) contamination due to atmospheric fallouts and industrial tailings originating from adjacent lead or zinc smelters. Three contrasting soils from two sampling sites were collected for the culture experiments.

Site 1 is called Prayon and belongs to the city of Trooz in the Liège Province. The two soils from this site were collected from a hill, which was intensively subjected to aerial dust and fumes emitted by the smoke stack from a former smelting plant dedicated to Zn production until 1974. The two soils were chosen according to their distinct substratum: a Givetian calcareous bedrock for the PC soil (for Prayon Calcareous soil) and an acid Famennian Shale bedrock for the PS soil (for Prayon Shale-derived soil). Due to the slight slope (20°), the chimney position and the main wind direction, the PC soil has been a preferential host of metal-bearing aerial fallouts. The Prayon site is intensively studied for its rich, diversified and specific flora associated with a spontaneous re-vegetation (Agrostis capillaris, Viola calaminaria, Noccaea caerulescens, Armeria maritima subsp. halleri). Site 2 is an industrial tailing located at Angleur, near the city of Liège, and between the Ourthe River and the Canal of the Ourthe. The slag heap built from waste deposits from a former Zn smelter that was operated from 1837 until 1905 by the “Société des mines et fonderies de zinc de la Vieille-Montagne” (for more details, see Ganne et al., 2006). This specific substrate favoured the establishment of adapted metallophyte flora (A. maritima subsp. halleri and A. capillaris). However, several parts of the slag-derived soil are still bare and one such bare soil was AB sampling soil site (for Angleur's Bare slag heap soil). In the present study, only soil samples from the surface layers (∼0–7 cm, Table 1) are considered. After collection on each site, several kilograms of soils were homogenized and air-dried in the laboratory, ready for the culture experiments.

For the culture experiment, two plant species were chosen on the basis of their high and fast biomass productivity, their suitability to the experiment conditions (previously tested by Chaignon and Hinsinger, 2003; Lambrechts et al., 2011), and their distinct physiological mechanisms: Brassica napus L. cv Adelie (rape) and Lolium multiflorum L. cv Meribel (ryegrass). L. multiflorum is a metal-tolerant monocotyledonous plant species, while B. napus belongs to the dicotyledonous species and is well known to be tolerant for some abiotic stresses, like high metal contents in soils (Houben et al., 2013; Marchiol et al., 2004). For L. multiflorum, no stem is observed, whereas a short stem is present for B. napus but was not isolated; only roots and shoots (including stems of B. napus) account for the chemical analyses in the present study.

The culture experiment was performed with a homemade cropping device whose design was adapted from Kruyts (2002) and Niebes et al. (1993) (Fig. 1), which allows us to sample roots with almost no soil contamination. Two PVC cylinders composed the cropping device: the plant compartment (internal diameter of 80 mm) and the soil compartment (internal diameter of 90 mm). The plants were germinated and grown on a polyamide mesh (mesh size: 20 μm) situated at the bottom of the plant compartment cylinder. The soil compartment lied on a cap covered by a second polyamide mesh (20 μm). All the apparatus parts were acid-washed (pH 3, HCl) and thoroughly rinsed with deionized water before assembly.

The experiment was conducted on 36 plant devices including three different substrates (three soil samples), two plant species, one nutrient solution and three replicates by treatment (control experiments have been performed in parallel, using clean sand as soil and deionized water as nutritive solution; due to the small amounts of dry plant parts in these control tests, elemental compositions have been measured—not reported here, but no isotopic ratio).

The plant devices were placed in a growth chamber where the following parameters were controlled: temperature (20 °C), humidity (around 90%) and light intensity (16-h photoperiod and mean light varying from 120 to 180 μmol m⁻²·s⁻¹, except during the seven first days of the hydroponic period where darkness was maintained). Two main growth periods were distinguished: the germination period (7 days) followed by the plant–soil contact period (14 days).

2.2.1 First step: germination phase

2.2.2 Second step: plant–soil contact

After sampling, soil samples were air-dried and sieved (2 mm). Following the methodology of Degryse et al. (2003), the soil Zn was extracted with CaCl₂ to estimate the isotope composition of the mobile fraction, i.e. the soil solution. This extract mimics the soil solution because the inorganic composition of the extract is similar to that of most soil solutions (Degryse et al., 2003). This method implies shaking of soil aliquots (2.5 g) in centrifuge tubes with 0.01 M CaCl₂ (25 ml) for 24 h in a lateral shaker. The samples were subsequently centrifuged (13,400 rpm for 5 min) and supernatants were filtered through 0.45-μm polycarbonate filters. The final solutions of CaCl₂ extracts were acidified before the ICP-AES analyses (Thermo Jarrell Ash Iris Advantage), which were performed at UCL (Belgium): Zn (and other metal) concentrations (ZnCaCl2) were measured (Table 1). Zinc concentrations in shoots, roots—as well as in the initial seeds—were also determined by ICP-AES after acid digestion of plant part powders in concentrated HNO₃ at 95 °C; the solution was subsequently evaporated to dryness and re-dissolved in a mixture of concentrated HNO₃ and HCl (1:3 ratio) at room temperature. Careful attention was paid to potential risks of Zn contamination by the reagents used for chemical preparations; procedural blanks were always used.

Whereas elemental analyses were performed on all the soil and plant parts, three replicates of each treatment (from the different combinations of plant species and soil samples) were carefully mixed and homogenized into a single composite sample for the isotopic analyses (after precise determination of Zn concentration and dry biomass, aliquots of equivalent Zn amounts of the three replicates were considered).

Chemical treatments of soil and plant samples and of CaCl₂ extracts were carried out under a class-100 laminar flow hood in a class-1000 cleaned room (at ULB). All the reagents used for sample treatments were distilled pro-analysis acids that were subsequently sub-boiled. The dilutions were performed with MilliQ water (18.2 MΩ·cm). Crushed soil and plant samples (about 2 mg) were previously dry ashed for 24 h at 450 °C in order to eliminate the organic matter. The samples for dry-ashing were dissolved by applying the tri-acid digestion technique (with concentrated 14 M HNO₃, 24 M HF and re-dissolution in 6 M HCl) using a Teflon Savillex^® beaker placed on a hot plate (120 °C) for evaporation until dryness. CaCl₂ extracts were simply evaporated. Zn was then purified by a novel chromatographic separation technique on micro-columns loaded with 0.2 ml of AG1-X8 resin and filled with successive additions of acids (HCl, HNO₃, HBr) (Mattielli et al., 2013). This method is especially adapted for the sample matrices characterized by high Zn concentrations compared to other cation abundances, allowing extremely low amounts of acids for elution and consequently low procedural blanks (≤2 ng of Zn). Upon separation, the eluate was dried down and digested with 100 μl of concentrated HNO₃ to dissolve the potential co-eluted organics. A full recovery of Zn was quantitatively monitored to circumvent problems associated with potential isotopic fractionation on columns. The yield values were higher than 98 ± 5% (values obtained from Zn purification of 10 plant and five soil samples). The total procedural blanks average ∼7 ng.

The Zn isotopic ratios were measured on an Nu plasma I MC-ICP-MS in wet plasma mode (ULB, Belgium). Zn (and Cu, used for the doping technique) isotopic compositions were measured by static multi-collection. A single analysis consisted of a measurement of 60 ratios, i.e. three blocks of 20 cycles with an integration interval of 10 s. On-peak baseline measurement with 30-s integration time prior to each analysis was done on a 0.05 M HNO₃ acid blank, which is then subtracted online during the analytical run of all the samples/standards. Nickel contributions were systematically corrected by monitoring mass 62 (⁶²Ni). Mass discrimination effects were corrected by using simultaneous external normalization (Cu-doping method) and standard-sample bracketing with in-house JMC Zn–Cu standard solutions (previously calibrated against the JMC 3-0749L Zn and NIST SRM 976 Cu reference standard solutions—for complementary information, see Mattielli et al. (2009) and Petit et al. (2008)). Every sample was run between two standards and was analysed at least in triplicate; the Zn isotopic composition was expressed in δ⁶⁶Zn relative to our in-house standard solution (Eq. (1))².

During the period of data acquisition, repeated measurements of in-house Zn and Cu standards show a mean δ⁶⁶Zn value of 0.00 ± 0.02‰ (2SD) (n = 117). In addition, repeated measurements of the Lyon JMC 3-0749L Zn standard solution batch used at ULB were performed and gave a mean δ⁶⁶Zn value of +0.11 ± 0.03‰ (2SD) (n = 17) relative to our in-house solution (0.00 ± 0.02‰). Consequently, the δ⁶⁶Zn results for the samples can then be converted relative to the JMC 3-0749L Zn by using the conventional conversion equation (Hoefs, 2008).

In order to assess the reproducibility, one sub-sample of soil horizons (AB) was digested in duplicate; the δ⁶⁶Zn values were +0.20 ± 0.04‰ and +0.26 ± 0.02‰ (the latter is reported in Table 3). In addition, to validate the chemical method of Zn separation, Zn from two soils (independent study) was isolated by another chromatographic technique using AG-MP1 resin and 0.5 M HNO₃ as an eluent (methodology described in Maréchal et al., 1999). The isotopic results obtained by the latter method (AG-MP1 resin) were −0.09 ± 0.01‰ and +0.23 ± 0.04‰, respectively for the first and second soil samples. The present method based on AG1-X8 resin gave the following δ⁶⁶Zn values: −0.05 ± 0.02‰ and +0.23 ± 0.04‰, for the first and second soil samples, respectively. Finally, to control the reproducibility of the entire analytical method and check the accuracy of the isotopic measurements on the MC-ICP-MS, the BCR281 (ryegrass) reference material was analysed and gave a δ⁶⁶Zn value of +0.38 ± 0.04‰ (n = 9), which is in good agreement with the previously published values (e.g., +0.40 ± 0.09‰ in Arnold et al., 2010b).

Zn concentration, dry biomass and transpiration volume data were systematically analysed using two-way ANOVAs, with the main factors being soils and plant species, followed by a multiple mean comparison test (Tukey's HSD test). The Spearman correlation coefficients (r_s) were calculated to determine the relationships between the isotopic and the elemental compositions. All statistical analyses were carried out using SAS 9.1 (SAS Institute, Cary, NC, USA).

Zinc is a major constituent in the soil samples, with abundances ranging from 7.4 g·kg⁻¹ in the shale-derived soil (PS) to 75 g·kg⁻¹ in the calcareous soil (PC). The soil sample significantly affected (P < 0.001) the dry plant biomass. Plants growing on the least Zn contaminated soils (PS) contained also the lowest Zn concentrations and had the highest biomass productions, suggesting that toxic Zn supply is a growth-limiting factor (Tables 1 and 2). The L. multiflorum species has produced significantly more root biomass than B. napus. The transpiration volume/device was statistically unaffected by soil sample or plant species. Total plant Zn concentrations (i.e. weighted average of shoot and roots) are 4 to 14 times less concentrated than in the corresponding soils. The Zn concentrations in B. napus roots (5.0 to 18.9 g·kg⁻¹) are significantly higher (P < 0.05) than in L. multiflorum roots (2.1 to 7.9 g·kg⁻¹). The Zn concentrations in seeds of B. napus and L. multiflorum were measured and represent a Zn input of only 0.05 mg and 0.09 mg per device, respectively. These values are well below the total Zn in biomass after culture, i.e. ≥1.95 mg for B. napus and 1.65 mg in plant per device for L. multiflorum. Soil sample and plant species have a significant effect (P < 0.05) on Zn concentrations in the whole plant as well as in roots (Table 2). In contrast, the shoot to root Zn concentration ratios vary between 0.12 and 0.42 and are only affected by the plant species, but not by the soil sample.

Prayon soils (PC and PS) are characterized by similar δ⁶⁶Zn values: −0.05 ± 0.02‰ (2SD) and −0.03 ± 0.01‰, respectively (δ⁶⁶Zn results relative to the JMC 3-0749L Zn standard solution; knowing that the average δ⁶⁶Zn value of the JMC 3-0749L standard gives +0.11 ± 0.03‰ (2SD) (n = 17) relative to our in-house solution (0.00 ± 0.02‰), meaning that δ⁶⁶Zn_in-house for PC and PS soils are +0.06 ± 0.02‰ and +0.08 ± 0.01‰, respectively). In contrast, Angleur soil (AB) shows a higher δ⁶⁶Zn of +0.26 ± 0.02‰ (Table 3). The CaCl₂ extracts systematically display lower δ⁶⁶Zn values compared to the bulk soils, indicating that the (surrogated) soil solutions are enriched in light Zn isotopes; factor Δ66ZnCaCl2−bulk soil³ reaches an almost constant value of −0.13‰ for the three soils (Table 3). As generally observed in plants (e.g., Houben et al., 2014; Tang et al., 2012; Weiss et al., 2005), δ⁶⁶Zn values in the shoots (δ⁶⁶Zn_shoot from −0.34 ± 0.06‰ to +0.06 ± 0.01‰, Table 3) are systematically lower relative to the roots (δ⁶⁶Zn_root from −0.10 ± 0.04‰ to +0.32 ± 0.04‰). But the isotopic fractionations during Zn translocation from roots to shoots are not constant; they are larger in B. napus (Δ⁶⁶Zn_shoot–root from −0.23 to −0.39‰) relative to L. multiflorum (Δ⁶⁶Zn_shoot–root between −0.04 and −0.20‰). The δ⁶⁶Zn values for the whole plant (Table 3) have been calculated by applying the mass balance equation:

Zinc isotopic signatures of the soils from the two distinct sites are typical values for contaminated soils (−0.05 ± 0.02‰ and −0.03 ± 0.01‰ for Prayon; +0.26 ± 0.02‰ for Angleur). In the literature, the published δ⁶⁶Zn data range from −0.2 to +1.5‰ (Bigalke et al., 2010; Dolgopolova et al., 2006; Fekiacova et al., 2015; Sivry et al., 2008; Sonke et al., 2008; Tang et al., 2012). Sonke et al. (2008) have demonstrated a remarkably homogeneous δ⁶⁶Zn value (+0.16 ± 0.20‰, n = 61 analyses for 10 mines) for ore-grade ZnS collected from various mining locations. Cloquet et al. (2008), Gioia et al. (2008) and Mattielli et al. (2009) have reported systematic light Zn isotope enrichments (relative to the initial ores) in aerial fallouts emitted by smoke stacks, which produce contaminated soils adjacent to smelters. Sivry et al. (2008) and Sonke et al. (2008) have demonstrated enrichments with heavy Zn isotopes in slag heaps, corresponding to solid residues of metallurgical process (smelting in blast furnace), complementary of aerial emissions. In agreement with those studies, the low positive δ⁶⁶Zn values of contaminated Prayon soils (PS and PC) mainly reflect the aerial fallouts from the former Zn smelter, while the significantly higher positive δ⁶⁶Zn values of the Angleur soil (AB) evidence the metallurgical scoria as the main Zn supply contributors in the soil.

The CaCl₂ extracts, i.e. the soil solution surrogates, are systematically enriched in light Zn isotopes relative to the corresponding bulk soil. A constant fractionation between soil and related soil solution (Δ66ZnCaCl2−bulk soil = −0.13‰) is found here irrespective of the soil sample. This confirms the initial observation of Arnold et al. (2010a,b) that Zn release from the soil solid phases to the liquid phase favours light Zn isotopes, which could reflect, according to Jouvin et al. (2009, 2012), a higher concentration of free Zn²⁺ ions in the soil solution.

The shoot Zn concentration ranges from 700 to 3300 mg Zn/kg. These values are at least one order above adequate Zn concentrations and reflect phytotoxicity due to Zn (Hamels et al., 2014). The Zn isotopic signature of the plants reflects that of the bulk soil. Within and among the plants, the isotope signatures differ. The Zn isotope signature in roots is isotopically heavier than in the corresponding soil solution (i.e. CaCl₂ extracts). Relatively comparable magnitudes of Zn isotopic fractionation between roots and soil solutions are observed for the two sites and the two plants, which suggests a common Zn transfer mechanism from the various soils to plants, irrespective of the Zn availability. In agreement with Weiss et al. (2005) and Jouvin et al. (2012), the enrichment with heavy isotopes in the roots relative to solution reflects an adsorption of heavy Zn isotopes from the soil solution onto root cell walls. Alternatively, as proposed by Aucour et al. (2011), the positive isotope fractionation associated with the roots might indicate that Zn is bound to a high-affinity ligand, with cellular sequestration.

It is generally accepted that the uptake of Zn from soil has at least four main mechanisms (see also Aucour et al., 2011; Caldelas et al., 2011; Houben et al., 2014; Jouvin et al., 2012; Tang et al., 2012 and references herein): (1) root-induced soil acidification and mobilization of soil Zn (Houben et al., 2014; Loosemore et al., 2004); (2) Zn adsorption on root cell wall (Hall, 2002); (3) non-specific Zn transfer through the plasma membrane into the root symplast via low-affinity transporters (Hacisalihoglu et al., 2001), or (4) Zn specific uptake by either ZIP proteins or phytosiderophores released by the plant roots to transfer the Fe/Zn complexes through the membrane (Claus et al., 2013). The present growth media are far from the Fe- or Zn-deficiency conditions, hence the release of siderophores by L. multiflorum roots was probably negligible (Römheld and Marschner, 1990). The B. napus (dicotyledonous) has a larger total plant Zn concentration compared to L. multiflorum (monocotyledonous), mainly because of drastically higher Zn concentrations in its roots. According to Dufey and Braun (1986), Dufey et al. (2001) and Meychik and Yermakov (2001), the Cationic Exchange Capacity of Roots (CECR, directly related to the negative charge density on roots) in dicotyledonous species is higher than in monocotyledonous species, enhancing Zn adsorption on root cell walls.

The isotopic signatures of Zn in the whole plant were almost identical for both plant species in each of the soils; however, differences among species emerged when considering either root or shoot (Fig. 2). In agreement with previous studies (Arnold et al., 2010a; Jouvin et al., 2009; Viers et al., 2007; Weiss et al., 2005), shoots systematically display lower δ⁶⁶Zn values compared to roots. As noticed by Caldelas et al. (2011), in Zn-excess conditions, roots systematically show positive δ⁶⁶Zn values relative to the solution reflecting enrichment in heavy Zn isotopes, whilst shoots are isotopically lighter and display negative or similar δ⁶⁶Zn values compared to the solution. There is more intense Zn isotope fractionation between roots and shoots in B. napus (Δ⁶⁶Zn_shoot–root from −0.23 to −0.39‰) than in L. multiflorum (Δ⁶⁶Zn_shoot–root from −0.04 to −0.20‰). Differences in Zn isotope fractionation between roots and shoots are correlated with the shoot–root Zn concentration ratios (Fig. 2), i.e. there is more enrichment in light Zn isotopes in shoots relative to roots as roots become more concentrated in Zn than shoots.

Isotope fractionation between roots and shoots might be the analogue of elemental fractionation, such as selective fractionation between strontium (Sr) and calcium (Ca) in plant parts, i.e. Sr accumulates in roots and Ca is more translocated to shoots, leading to a decreasing Sr:Ca ratio in plants with increasing distance from the roots (Drouet and Herbauts, 2008; Smolders and Merckx, 1993). Like Sr, in our experiment, heavy Zn isotopes accumulate in the root system and light Zn isotopes are preferentially transported further through the plant.

It is generally assumed that three main mechanisms are involved in the Zn transport from roots to shoots of plants: (1) exchange on the sorption sites of the plant tissues (adsorption on cell walls from the roots but also in the xylem vessel) and Zn loading into the xylem driven by heavy-metal ATPase transporting (Hanikenne et al., 2008); (2) diffusion (Zn transport on relatively short distance under concentration gradient) (Barberon and Geldner, 2014; Caldelas et al., 2011; Moynier et al., 2009; Weiss et al., 2005 and references herein); and (3) convection (mass flow transport in plant controlled by transpiration) (Barberon and Geldner, 2014; Lorenz et al., 1994). So far, the two first processes were essentially the only ones explored with the isotopic tool: as previously described by Jouvin et al. (2012), metal transport is carried out by the metal complexation with organic acids or amino acids or peptides produced by the plant. Those metals cross subsequently the cell membranes and circulate through the channels between cells and along the xylem and phloem by diffusion and ion exchange reactions, which induce Zn isotope fractionation. According to Rodushkin et al. (2004), Zn isotope fractionation occurring during diffusive transport would preferentially favour the transport of light Zn isotopes, and an increase of the fractionation extent with the diffusive distance. Several previous studies corroborated the hypothesis of long-distance diffusion process accounting for Zn isotope fractionation during transport in trees, palm trees, and herbaceous species (Viers et al., 2007), in bamboos compared to lentils (Moynier et al., 2009) and in Phragmites australis (Caldelas et al., 2011) in Zn-deficient or Zn-sufficient conditions. Those authors hypothesized a relationship between the fractionation extent and the length of the pathway or height of the plants. In the present work, in Zn-excess conditions all plant materials are very small, with a similar height of ∼10 cm, but the two species differ in shoot and root isotopic compositions and display large Δ⁶⁶Zn_shoot–root (up to −0.39 ± 0.07‰). This suggests the control of Zn fractionation during translocation by another process, potentially species-dependent.

By attempting to confirm or refute the diffusion mechanism hypothesis, three simple diffusion models can be tested. A first simple calculation based on a one-dimensional diffusion model was conducted. If the plant is considered as a solid in contact with a liquid solution characterized by a constant Zn concentration, Zn will diffuse from the solution to the inside of the plant. A diffusion front will steadily advance inward the plant over a distance that can be roughly estimated to be the square root of D × t, where D is the diffusion coefficient with a value of 5·10⁻⁶ cm²·s⁻¹ (approximation of the diffusion coefficient for Zn²⁺ in a free solution; Rodushkin et al., 2004) and t is the time. In the present study, the experiment duration t of 14 days corresponds to a distance reached by the diffusion front of only 2.46 cm, which is incompatible with the observed fractionation throughout the plant standing ±10 cm height. Second, assuming Δ⁶⁶Zn = δ⁶⁶Zn_top − δ⁶⁶Zn_entrance, the plant isotopic fractionation due to diffusion process can be also estimated by the following equation, according to Moynier et al. (2009):

The shoot–root Zn isotope fractionation exhibits a marked strong negative correlation with transpiration volume per dry biomass unit across species and soil samples (Fig. 3). This suggests a transpiration-controlled isotope discrimination. In disagreement with Aucour et al. (2011), who postulated that advection with transpiration stream does not fractionate the Zn isotopes, our results show higher Δ⁶⁶Zn_shoot–root when the transpiration volume per biomass unit increases. Active transport (through carrier proteins) is associated with plants in metal-deficient conditions (or low metal concentrations) (Barberon and Geldner, 2014), while passive transport (transpiration flow) is generally observed in plant–soil systems displaying high metal levels (Hopmans and Bristow, 2002). Along the same lines, diffusion processes in soil or in the rhizosphere may control Zn uptake under deficient conditions, whereas mass flow may control the flux of Zn from soil to plants under high Zn supply (Degryse et al., 2009). Consequently, the passive Zn translocation from roots to leaves (or shoots) follows the transpiration path (a convection mechanism) within the xylem (Broadley et al., 2007; Page and Feller, 2005; Page et al., 2006a,b), and the effect of protein carriers would be negligible compared to the transpiration flow. The passive transport hypothesis by mass flow of water was also postulated by Henriet et al. (2006) for Si transport in banana at high Si concentrations.

It is puzzling how the transpiration flux on a plant dry weight affects the isotope discrimination between roots and shoots as shown in Fig. 3. The observed trend is weighted by the differences between both plants. Both plants have similar total plant Zn (μg Zn/plant) when compared pairwise per soil and have even similar shoot:root total Zn partition, albeit the B. napus plants have much smaller roots with higher Zn concentrations than L. multiflorum. As a result, the total flux of Zn from soil to plants is almost identical for each plant (per soil); however, there is more water flux on a total plant weight basis for rape than for ryegrass, i.e. a higher water flow across roots (weight based) for B. napus compared to L. multiflorum. One might suggest that Zn forms control fractionation along the soil–root–xylem path: free Zn could be transported faster towards the shoots than Zn-organic complexes that preferentially bind heavy Zn isotopes (Jouvin et al., 2009, 2012). As water flow increases per unit root weight, so increases the isotope fractionation. As a result, in the context of a highly contaminated soil–plant system, our data suggest that the translocation of Zn is mainly driven by the plant transpiration flow implying a convection mechanism rather than a diffusive process. During the translocation from roots to shoots, heavy isotopes are preferentially adsorbed on negatively charged sites of the cell walls from the surrounding xylem tissues, whereas light isotopes remain preferentially in the xylem sap; the proportion of light isotopes in the xylem sap increasing with increasing the transpiration volume/dry biomass ratio.