Zinc in plants

1. Zinc inputs to soils

The primary input of Zn to soils is from the chemical and physical weathering of parent rocks. The lithosphere typically comprises 70–80 µg Zn g⁻¹, whilst sedimentary rocks contain 10–120 µg Zn g⁻¹ (Friedland, 1990; Barak & Helmke, 1993; Alloway, 1995). Mean soil Zn concentrations ([Zn]_soil) of 50 and 66 µg total Zn g⁻¹ soil are typical for mineral and organic soils, respectively, with most agricultural soils containing 10–300 µg Zn g⁻¹ (Alloway, 1995; Barber, 1995). Zinc occurs in rock-forming minerals as a result of the nonspecific replacement of Mg and Fe with Zn (Barak & Helmke, 1993). Rocks containing weathered Zn minerals, including Zn sulphide (sphalerite, wurtzite), sulphate (zincosite, goslarite), oxide (zincinte, franklinite, gahnite), carbonate (smithsonite), phosphate (hopeite) and silicate (hemimorphite, willemite) minerals, can form ‘calamine’ soils containing extremely high concentrations of Zn and other metals (Barak & Helmke, 1993). For example, in Plombières in Belgium, [Zn]_soil exceeds 100 000 µg Zn g⁻¹ (Cappuyns et al., 2006). Such sites are usually localized to a few hectares, although adjacent soils can also have high [Zn]_soil through water seepage from ore bodies (Chaney, 1993). Secondary natural inputs of Zn to soils arise because of atmospheric (e.g. volcanoes, forest fires, and surface dusts) and biotic (e.g. decomposition, leaching/washoff from leaf surfaces) processes (Friedland, 1990).

Humans have long influenced Zn inputs to soils. Two thousand years ago, approx. 10 000 tonnes Zn yr⁻¹ were emitted as a result of mining and smelting activities (Nriagu, 1996). Since 1850, emissions have increased 10-fold, peaking at 3.4 Mt Zn yr⁻¹ in the early 1980s, and then declining to 2.7 Mt Zn yr⁻¹ by the early 1990s (Nriagu, 1996). Arctic troposphere Zn concentrations (c. 2 ng Zn m⁻³ in winter months) are yet to reflect this decline (Gong & Barrie, 2005). The ratio of Zn emissions arising from anthropogenic and natural inputs is estimated to be > 20 : 1 (Friedland, 1990). Other anthropogenic inputs of Zn to soils include fossil fuel combustion, mine waste, phosphatic fertilizers (typically 50–1450 µg Zn g⁻¹), limestone (10–450 µg Zn g⁻¹), manure (15–250 µg Zn g⁻¹), sewage sludge (91–49 000 µg Zn g⁻¹), other agrochemicals, particles from galvanized (Zn-plated) surfaces and rubber mulches (Chaney, 1993; Alloway, 1995). Crop Zn toxicity can occur in Zn-contaminated soils (discussed in Section VI.1).

2. Zinc behaviour in soils

Soil Zn occurs in three primary fractions: (i) water-soluble Zn (including Zn²⁺ and soluble organic fractions); (ii) adsorbed and exchangeable Zn in the colloidal fraction (associated with clay particles, humic compounds and Al and Fe hydroxides); and (iii) insoluble Zn complexes and minerals (reviewed by Lindsay, 1979; Barrow, 1993; Alloway, 1995; Barber, 1995). The distribution of Zn between soil fractions is determined by soil-specific precipitation, complexation and adsorption reactions. The dominant factor determining soil Zn distribution is pH; Zn is more readily adsorbed on cation exchange sites at higher pH and adsorbed Zn is more readily displaced by CaCl₂ at lower pH. Thus, soluble Zn and the ratio of Zn²⁺ to organic Zn-ligand complexes increase at low pH, especially in soils of low soluble organic matter content. Soil type, soil moisture, mineral and clay types and contents, diffusion and mass flow rates, weathering rates, soil organic matter, soil biota and plant uptake will also affect Zn distribution. Insoluble Zn comprises > 90% of soil Zn and is unavailable for plant uptake. Exchangeable Zn typically ranges from 0.1 to 2 µg Zn g⁻¹. Concentrations of water-soluble Zn in the bulk soil solution ([Zn]_bss) are low, typically between 4 × 10⁻¹⁰ and 4 × 10⁻⁶ m (Barber, 1995), even in Zn-contaminated soils (Knight et al., 1997). Numerous Zn-ligand complexes can exist in solution which can be difficult to measure directly, and speciation models, based on total dissolved concentrations of elements and ligands, their stability constants and mineral equilibria reactions, are often used to infer [Zn²⁺]_bss (Barak & Helmke, 1993; Zhang & Young, 2006). Zn²⁺ typically accounts for up to 50% of the soluble Zn fraction and is the dominant plant-available Zn fraction. However, in calcareous soils, Zn²⁺ may be as low as 10⁻¹¹−10⁻⁹ m and can limit crop growth (Hacisalihoglu & Kochian, 2003; discussed in Section V.1).

Soil Zn fractions in the solid phase can be quantified using sequential extractions or isotopic dilution techniques (Young et al., 2006). For example, Zn extracted by H₂O (‘water-soluble’), KNO₃ (‘exchangeable’), Na₄P₂O₇ (‘organically bound’), EDTA (‘carbonate/noncrystalline iron occluded’), NH₂OH (‘manganese oxide occluded’), Na₂S₂O₄ (‘crystalline iron oxide occluded’), HNO₃ (‘sulphides’), HNO₃ + H₂O₂ (‘residual’) represented 0.2, 10.0, 32.5, 7.9, 7.2, 7.5, 3.3 and 32.3% of total soil Zn, respectively, in a mineral soil from Indiana, in the USA (Miller & McFee, 1983). Despite recent methodological advances in measuring metal species in the soil solution (Zhang & Young, 2006), Zn availability at the soil–root interface ([Zn]_ext) can still be difficult to determine satisfactorily, especially if Zn uptake by roots is high and Zn-depletion zones develop around the root. Thus, modelling approaches to determine soil and plant effects on Zn dynamics have been developed (Barber & Claassen, 1977; Bar-Yosef et al., 1980; Barber, 1995; Whiting et al., 2003; Qian et al., 2005; Lehto et al., 2006). For example, Whiting et al. (2003) adapted a transport model of Baldwin et al. (1973) to calculate [Zn]_ext, using empirically derived soil parameters (Barber, 1995) and empirically derived or inferred plant parameters. The parameters were [Zn]_bss (concentration of Zn in the bulk soil solution), D (the effective diffusion coefficient of Zn in the soil solution), b (the Zn-buffering power of the soil), x (the radius of the soil cylinder that can be exploited by the root), α (the root absorption power of plant for Zn), õ (the water flux from the soil to the root surface) and a (the root radius). Values for [Zn]_bss vary widely, as described previously, typically in the range 1 × 10⁻⁸ to 1 × 10⁻⁶m. Values of D also vary widely, increasing at high [Zn]_soil, high soil bulk density (air-filled pores increase diffusion resistance), high soil water content and low pH, typically within the range 10⁻¹⁰−10⁻⁸ cm² s⁻¹ for Zn (Barber, 1995). Values of b represent the distribution of Zn between the solution and the solid phases and are estimated from empirically derived Zn adsorption isotherms (Barber, 1995). Values of b are highest at low [Zn]_bss, and thus at high cation exchange capacity (CEC) and pH. Values of b typically range from 2.4 to 571 (Barber, 1995), and for vertisols from 217 to 790 (Dang et al., 1994). Paramaters õ and α represent plant-specific transpiration and Zn uptake rates and root morphology. For plants of the same size and with the same root Zn and water-absorbing power, the primary drivers of [Zn]_ext are [Zn]_bss and b (Whiting et al., 2003). At high b (> 200), [Zn]_ext is proportional to, or approximates, [Zn]_bss, that is, plant-available Zn is determined by the capacity of the soil to replenish the soluble Zn fraction as it is removed by the plant. However, when b < 200, [Zn]_ext is << [Zn]_bss, and especially so when b < 50.

Modelling [Zn]_ext requires assumptions to minimize complexity and to compensate for a lack of adequate input data. For example, Whiting et al. (2003) assumed that: (i) [Zn]_ext was in steady state with roots nonrandomly dispersed at constant density in a finite volume of soil; (ii) only radial transport of Zn occurred; (iii) all Zn in the soil solution was available for uptake; and (iv) no compounds mobilizing Zn from the solid phase were secreted from roots, mycorrhiza or other soil-dwelling organisms. However, models to predict [Zn]_ext are continually being improved for both soil (Lehto et al., 2006) and plant (Qian et al., 2005) parameters, and are likely to become invaluable in improving our understanding of whole-plant physiological processes, optimizing crop Zn nutrition through the application of fertilizers and soil amendments, and improving risk assessments for metal-contaminated environments.

3. Zinc fluxes into plants

Zinc is acquired from the soil solution primarily as Zn²⁺, but also potentially complexed with organic ligands, by roots which feed the shoots via the xylem. The relationship between Zn influx (and uptake or accumulation) to excised roots and intact plants, V, and [Zn]_ext is often characterized by the sum of one or more Michaelis–Menten functions, each defined by a V_max (the rate at [Zn]_ext = ∞) and an affinity constant, K_m ([Zn]_ext when V = 0.5V_max), plus a linear term, k (V/[Zn]_ext). Most detailed kinetic studies report a Michaelis–Menten function with a K_m of 1.5–50 µm (with V_max values of up to 5.74 µmol g⁻¹ FW h⁻¹), and, occasionally, additional Michaelis–Menten functions with higher K_m values (Table 2). These K_m values are generally higher than [Zn²⁺]_bss (10⁻¹¹−10⁻⁶ m; see Section IV.2). Two studies have reported Michaelis–Menten functions for wheat plants with K_m values of 0.6–2.3 nm for Zn²⁺ uptake (Wheal & Rengel, 1997; Hacisalihoglu et al., 2001). Whilst K_m and V_max do not differ greatly between Zn-efficient and Zn-inefficient wheat varieties, suggesting that the kinetics of Zn influx per se do not play a significant role in Zn efficiency in wheat (Hart et al., 1998, 2002; Hacisalihoglu et al., 2001; Hacisalihoglu & Kochian, 2003), differences in these parameters between Zn-efficient and Zn-inefficient genotypes of sugarcane, rice and tomato have been reported (Bowen, 1973, 1986, 1987). In general, Zn-deficient plants have higher V_max and comparable K_m values to Zn-replete plants (Rengel & Wheal, 1997; Hacisalihoglu et al., 2001). Several authors have attributed a linear term, which is not removed by washing the roots in solutions containing excess cations, to the accumulation of Zn bound strongly to cell walls (Lasat et al., 1996; Hart et al., 1998, 2002; Hacisalihoglu et al., 2001). There has been some discourse about the consequences of Zn binding to cell walls in experiments to determine kinetic parameters for Zn influx to roots (Reid et al., 1996). In the giant alga, Chara, cell walls can be removed and apoplastic effects on Zn influx can be determined directly (Reid et al., 1996). Using this technique, Zn influx to the cytoplasm was described by the sum of a Michaelis–Menten function with a low K_m (< 0.1 µm) plus a linear term for [Zn]_ext up to 50 µm. These data suggest that most experiments to determine the kinetic parameters of Zn influx to roots may have overestimated K_m values and distorted the linear term.

Transport from epidermal and cortical cells to the root xylem can occur via the cytoplasmic continuum of cells, linked by plasmodesmata, from which Zn is pumped into the stelar apoplast (Lasat & Kochian, 2000). This ‘symplastic’ pathway is catalysed by plasma membrane and tonoplast transport activity. Zinc can also be delivered extracellularly to the stelar apoplast in regions where the Casparian band is not fully formed (White, 2001; White et al., 2002b). Apoplastic mineral fluxes are dominated by the cell wall cation exchange capacity (CEC), Casparian band formation and water flows (Sattelmacher, 2001). Both symplastic and apoplastic fluxes may contribute to net Zn fluxes to the shoot. If symplastic fluxes dominate, Zn accumulation in the shoot approximates the sum of unidirectional influx (φ_oc) of Zn to root cells, minus efflux (φ_co) and net vacuolar Zn sequesteration (i.e. unidirectional vacuolar influx, φ_cv, minus unidirectional vacuolar efflux, φ_vc) (Lasat & Kochian, 2000; White et al., 2002b). Thus, the maximal [Zn]_shoot supported by symplastic root Zn fluxes can be simulated for different shoot : root FW ratios, relative growth rates (RGRs) and φ_oc values if one assumes steady-state conditions (White et al., 2002b). In Thlaspi caerulescens J. & C. Presl., which can accumulate > 30 000 µg Zn g⁻¹ DW (see Section VI.3), the delivery of Zn to the root xylem exclusively via the symplast is kinetically challenging at high [Zn]_ext (White et al., 2002b). Substantial symplastic Zn fluxes may occur at high [Zn]_ext (White et al., 2002b). However, it is noteworthy that the Ca²⁺-transporting ATPase activity required to catalyze the export of Ca²⁺ alone from the symplast into the root stelar apoplast may exceed total membrane protein quotas in a typical dicot (White, 1998; White et al., 2002b). Thlaspi caerulescens has substantial shoot Ca, Mg and Zn concentrations (> 60,000 µg g⁻¹ DW; Meerts et al., 2003; Dechamps et al., 2005). Thlaspi caerulescens also has altered Casparian band formation (Section VI.iii), and Cd²⁺ fluxes are substantially greater at the root tips in wheat, T. caerulescens and T. arvense L. (i.e. in regions where the Casparian band is not fully formed; Piñeros et al., 1998). Integrating Zn flux analyses with models to predict [Zn]_ext may ultimately enable quantification of the relative symplastic and apoplastic root fluxes, provided fluxes and intracellular Zn distibutions of intact root cells can be determined accurately.

1. Zinc is an essential plant nutrient

The essentiality of Zn in plants was first shown in maize (Mazé, 1915), and subsequently in barley and dwarf sunflower (Sommer & Lipman, 1926). Early reports of severe Zn deficiency symptoms included impaired stem elongation in tomato (Skoog, 1940). Incipient Zn deficiency symptoms in tomato, remedied by resupply of Zn, included reduced protein and starch synthesis whilst sugar content was unaffected (Hoagland, 1948). Severe Zn deficiency symptoms have since been catalogued for many crops (Scaife & Turner, 1983; Marschner, 1995; Sharma, 2006; Fig. 1). Severe Zn deficiency is characterized by root apex necrosis (‘dieback’), whilst sublethal Zn deficiency induces spatially heterogeneous or interveinal chlorosis (‘mottle leaf’), the development of reddish-brown or bronze tints (‘bronzing’), and a range of auxin deficiency-like responses such as internode shortening (‘rosetting’), epinasty, inward curling of leaf lamina (‘goblet’ leaves) and reductions in leaf size (‘little leaf’). In most crops, the typical leaf Zn concentration ([Zn]_leaf) required for adequate growth approximates 15–20 mg Zn kg⁻¹ DW (Marschner, 1995).

Zinc is the most common crop micronutrient deficiency, particularly in high-pH soils with low [Zn]_bss (Graham et al., 1992; White & Zasoski, 1999; Cakmak, 2002, 2004; Alloway, 2004). Notably, 50% of cultivated soils in India and Turkey, a third of cultivated soils in China, and most soils in Western Australia are classed as Zn-deficient. Zinc deficiency in crop production can be ameliorated through agronomy or genetic improvement. Early agronomic successes included the treatment of little leaf in peach orchards, using soil-applied FeSO₄ fertilizers containing Zn impurities, and subsequently the treatment of mottle leaf in citrus orchards, rosetting in pecan and stunted pineapple growth using foliar sprays containing Zn (Hoagland, 1948). More recently, substantial arable crop responses to Zn fertilization have been reported in Australia, India, and Central Anatolia in Turkey, where wheat grain yields have increased by over 600% since the mid-1990s, with a concomitant annual economic benefit of US$100 million (Cakmak, 2004). It may also be possible to improve crop yields on Zn-deficient soils by exploiting genotypic differences in Zn uptake and tissue-use efficiency that exist within crop species (Rengel, 2001; Cakmak, 2002; Hacisalihoglu & Kochian, 2003; Alloway, 2004). For example, in rice, Zn uptake efficiency correlates with exudation rates of low-molecular-weight organic anions and a substantial proportion of the phenotypic variation in Zn uptake efficiency is under genetic control (Hoffland et al., 2006; Wissuwa et al., 2006).

2. Evolutionary aspects of shoot Zn concentration among angiosperms

Previously, a meta-analysis of 70 species from 35 studies in the literature implied that ancient evolutionary processes might impact on [Zn]_shoot in angiosperms (Broadley et al., 2001). A much larger survey was undertaken here. Primary [Zn]_shoot data, that is, Zn concentrations reported in leaf or nonwoody shoot tissues, were obtained from 1108 studies, contained in 204 published papers and three unpublished datasets (Tables S6–S9). These primary data define the largest set of interlinked studies, in which [Zn]_shoot values are reported for more than two species in controlled experiments where each study contains more than one species common to another study. Thus, 365 species from 48 families and 12 key clades were sampled. Primary [Zn]_shoot data were ln-transformed and a variance-components model was fitted to the data using residual maximum likelihood (REML) procedures (Broadley et al., 2001, 2003). Species, family and key clade variance components were estimated using a random model factor of (study + (key clade/family/species)) with no fixed model factor. Mean data for phylogenetic groups and significance tests (Wald tests) were conducted using (key clade/family/species) as a fixed factor, retaining (study + (key clade/family/species)) as a random factor. All analyses were performed using GenStat (Release 9.1.0.147; VSN International, Oxford, UK).

Study effects dominated the variance components because of the vast range of [Zn]_ext across all studies and because of differences in the relationship between [Zn]_ext and [Zn]_shoot between species (Table S9). After removing study effects, key clade (Wald statistic = 592.15, d.f. = 10, P < 0.001) and family within key clade (Wald statistic = 298.71, d.f. = 36, P < 0.001), variance components accounted for 22.1% of the variation in [Zn]_shoot among angiosperm species (Table S9). The analysis was repeated with Thlaspi and Arabidopsis genera excluded to remove the influence of Zn ‘hyperaccumulation’ (defined in Section VI.3) from the analysis. Adjusted key clade and family variance components accounted for 26.5% of the variation in [Zn]_shoot (Table S9). After removing the effect of study, and with Thlaspi and Arabidopsis genera excluded, mean relative [Zn]_shoot among key clades ranged from < 40 (Ericales) to > 100 (Caryophyllales and noncommelinoid monocotyledons; NB: units approximate µg Zn g⁻¹ DW) (Table S9). Among well-replicated families, lower [Zn]_shoot occurred in Linaceae (52, n = 29), Poaceae (59, n = 1527) and Solanaceae (66, n = 181), and higher [Zn]_shoot occurred in Amaranthaceae (108, n = 214) and Salicaceae (195, n = 45) (Table S9). Compared with other essential elements, variation in [Zn]_shoot manifesting above the family level is less than for Ca, K, Mg and Si (Broadley et al., 2004; Hodson et al., 2005), but greater than for N and P (Broadley et al., 2004). Furthermore, variation in [Zn]_shoot manifesting above the family level is less than previously reported for [Zn]_shoot among a limited dataset of 70 species (Broadley et al., 2001). However, despite the large study effects within this dataset, evidence of ancient evolutionary processes influencing [Zn]_shoot can still be observed using this approach. A considerable additional sampling effort is required to explore this phenomenon further.

After removing study effects, > 70% of the remaining variation in [Zn]_shoot occurs within-family. Thus, substantial differences in [Zn]_shoot exist between and within genera and species. Several billion people worldwide have Zn-deficient diets, and this within-species genetic variation is providing a valuable genetic resource to select or breed crops with increased Zn concentrations in their edible portions, notably in several cereal, legume, root and leafy vegetable crops (Graham et al., 2001; Welch & Graham, 2004; Grusak & Cakmak, 2005; Vreugdenhil et al., 2005; White & Broadley, 2005; Ghandilyan et al., 2006). Accelerated breeding to increase the delivery of dietary Zn in crops may be possible following identification of intraspecific genetic variation in Zn composition. In common bean (Phaseolus vulgaris L.), seed Zn concentration ([Zn]_seed) is a quantitative trait which can be mapped to genetic loci using quantitative trait loci (QTL) analyses (Ghandilyan et al., 2006). QTLs for [Zn]_seed have also been mapped in A. thaliana (Vreugdenhil et al., 2004). Following the identification of QTL, candidate genes or loci can be resolved through fine mapping and map-based cloning, and this information could be used for gene-based selection or marker-assisted breeding strategies. An advantage of this strategy is that knowledge of the genes and/or chromosomal loci controlling [Zn]_seed or [Zn]_shoot in one plant species could be used in a different target crop species by exploiting gene homology and/or genome collinearity (Ghandilyan et al., 2006).

1. Zn toxicity in crops

Zinc toxicity in crops is far less widespread than Zn deficiency. However, Zn toxicity occurs in soils contaminated by mining and smelting activities, in agricultural soils treated with sewage sludge, and in urban and peri-urban soils enriched by anthropogenic inputs of Zn, especially in low-pH soils (Chaney, 1993). Toxicity symptoms usually become visible at [Zn]_leaf > 300 mg Zn kg⁻¹ leaf DW, although some crops show toxicity symptoms at [Zn]_leaf < 100 mg Zn kg⁻¹ DW (Chaney, 1993; Marschner, 1995), and toxicity thresholds can be highly variable even within the same species. For example, [Zn]_leaf associated with a 50% yield reduction in radish ranged from 36 to 1013 mg kg⁻¹ DW (Davies, 1993). Zn toxicity symptoms include reduced yields and stunted growth, Fe-deficiency-induced chlorosis through reductions in chlorophyll synthesis and chloroplast degradation, and interference with P (and Mg and Mn) uptake (Carroll & Loneragan, 1968; Boawn & Rasmussen, 1971; Foy et al., 1978; Chaney, 1993). Crops differ markedly in their susceptibility to Zn toxicity. In acid soils, graminaceous species are generally less sensitive to Zn toxicity than most dicots, although this is reversed in alkaline soils (Chaney, 1993). Among dicots, leafy vegetable crops are sensitive to Zn toxicity, especially spinach and beet, because of their inherent high Zn uptake capacity (Boawn & Rasmussen, 1971; Chaney, 1993). There is also genetic variation in sensitivity to Zn toxicity within species, including soybean (Glycine max L.; Earley, 1943; White et al., 1979a,b,c) and rice (Oryza sativa L.), in which QTLs impacting on sensitivity to Zn toxicity have recently been mapped (Dong et al., 2006).

2. Plant tolerance to elevated soil Zn (hypertolerance)

Numerous species of metallophyte thrive at a [Zn]_soil that is toxic to most crop plants. Early studies on Zn hypertolerance focused on elevated [Zn]_soil at mine sites and/or near corroded galvanized materials, such as electricity pylons (reviewed by Antonovics et al., 1971; Baker, 1987; Ernst et al., 1990, 1992; Macnair, 1993). Initially, Zn hypertolerance was thought to occur most frequently in species of Poaceae, Caryophyllaceae and Lamiaceae, although subsequent studies have shown no phylogenetic predisposition to the evolution of Zn hypertolerance (Baker, 1987). However, this hypothesis is difficult to test directly. At a species level, elegant theoretical and experimental frameworks have been developed to characterize the genetics of Zn hypertolerance (Macnair, 1990, 1993). Briefly, Zn hypertolerance tends to manifest as a dominant phenotype and spread rapidly in a population. However, dominance is not a fixed property of a gene but a measure of the phenotypic deviation of the heterozygote from the mean of the two homozygotes. Thus, G × E interactions will affect whether a gene confers a dominant or recessive phenotype under any particular condition, rendering the study of Zn tolerance nontrivial (Macnair, 1990, 1993). Nevertheless, Zn hypertolerance is likely to be under the control of a small number of major genes, as in Silene vulgaris (Moench) Garcke (Schat et al., 1996) and Arabidopsis halleri (L.) O’Kane & Al-Shehbaz (Macnair et al., 1999). In addition to dominance (i.e. interallelic/intragenic) effects, epistatic (i.e. intergenic) interactions have an impact on Zn hypertolerance, although these effects are more difficult to quantify (Macnair, 1993). Zinc-hypertolerant plants show fitness costs at low [Zn]_soil, for example, suboptimal carbonic anyhdrase and nitrate reductase activity occurs in S. vulgaris (Ernst et al., 1992). In Silene dioica (L.) Clairv., pollen selection may accelerate reproductive isolation between adjacent populations which differ in Zn tolerance (Searcy & Mulcahy, 1985a,b,c).

At a whole-plant scale, natural Zn hypertolerance is thought to be conferred by Zn exclusion or by compartmentalization, for example, through mycorrhizal symbioses, altered root-to-shoot translocation or accumulation in older leaves (Baker, 1981; Ernst et al., 1992; Hall, 2002). Key subcellular processes enabling Zn hypertolerance are likely to be increased organic acid production and vacuolar compartmentalization, including increased Zn efflux across the plasma membrane (Ernst et al., 1992; Verkleij et al., 1998; Chardonnens et al., 1999; Clemens, 2001; Hall, 2002). Early hypotheses that cell-wall Zn binding, reduced membrane leakage or increased phytochelatin production increased Zn tolerance have not been supported by subsequent studies (Ernst et al., 1992; Harmens et al., 1993; Clemens, 2001; Schat et al., 2002).

In general, Zn hypertolerance does not segregate with other metal tolerance phenotypes, although Cd, Co, Cu, Cd, Ni and/or Pb cotolerance can occur (Gregory & Bradshaw, 1965; Wu & Antonovics, 1975; Cox & Hutchinson, 1980; Symeonidis et al., 1985; Macnair, 1990, 1993; Brown & Brinkmann, 1992; Schat & Vooijs, 1997). Cotolerance can be either pleiotropic (i.e. strict cotolerance) or the result of multiple hypertolerance mechanisms. Multiple hypertolerances arise through nonrandom association of favourable alleles at two or more loci (i.e. linkage disequilibrium), potentially in response to the co-occurrence of several metals in soils, gene flow from adjacent sites or seed transport between different metalliferous sites (Macnair, 1993; Schat & Vooijs, 1997). For example, Zn/Co/Ni cotolerance phenotypes cosegregated in Silene vulgaris, implying a pleiotropic effect, whilst Zn/Cd/Cu cotolerance phenotypes segregated independently, implying multiple mechanisms (Schat & Vooijs, 1997). Zinc/Cu cotolerance phenotypes also segregated independently in populations of Agrostis stolonifera L. (Wu & Antonovics, 1975) and Mimulus guttatus DC. (Macnair, 1990). Zinc-hypertolerant cultivars of grass species have been successfully bred and used to revegetate soils contaminated with Zn, Pb and other heavy metals, following mining activities, including Festuca rubra L. cv. Merlin and Agrostis capillaris L. cv. Groginan for calcareous and acid wastes, respectively (Smith & Bradshaw, 1979; Whiting et al., 2005).

3. Zinc hyperaccumulation

The ability of some Zn-hypertolerant metallophytes to accumulate exceptional concentrations of Zn in their aerial parts was probably first reported among ‘galmei’ (calamine) flora of the Aachen region on the border of Belgium and Germany, where the presence of > 1% Zn in the plant ash of Viola calaminaria (Ging.) Lej. was recorded in 1855 (Reeves & Baker, 2000). A [Zn]_shoot > 1% DW in Thlaspi alpestre L. (= T. caerulescens) was reported shortly thereafter in 1865 (Reeves & Baker, 2000). Zinc hyperaccumulation has since been defined as the occurance of > 10 000 µg Zn g⁻¹ DW in the aerial parts of a plant species when growing in its natural environment (Baker & Brooks, 1989). These species (Table 3) have come to dominate the literature, in part because of the desire to transfer the character into crop species for use in phytoremediation, phytomining and biofortification (Chaney, 1983; Baker & Brooks, 1989; Brooks, 1998; Salt et al., 1998; Guerinot & Salt, 2001; Macnair, 2003; Krämer, 2005a). Typically, 10–20 species are reported to be Zn hyperaccumulators, with a smaller number of these able to accumulate Cd to very high concentrations as well. This raises the intriguing question: how often has the Zn hyperaccumulation character evolved? The answer is uncertain because: (i) 3000 µg Zn g⁻¹ DW might be a more suitable evolutionary definition of extreme [Zn]_shoot based on the frequency distribution of values observed within the genus Thlaspi s.l., which contains most Zn hyperaccumulators (Reeves & Brooks, 1983; Reeves & Baker, 2000); (ii) records for several nonbrassicaceous Zn hyperaccumulators are uncertain (Macnair, 2003; R. D. Reeves, pers. comm.) (NB: plant samples collected from metal-rich substrates can easily become contaminated by soil particles); and (iii) taxonomic uncertainties and the use of synonyms can affect the identification of field and herbarium samples.

The genus Thlaspi L. s.l. is likely to be polyphyletic and its taxonomy is controversial. Seed-coat anatomical and sequence data have thus been used to split the genus into several alternative genera, including Thlaspi s.s., Vania and a clade containing Thlaspiceras, Noccaea, Raparia, Microthlaspi and Neurotropis (Meyer, 1973, 1979; Mummenhoff & Koch, 1994; Mummenhoff et al., 1997; Koch & Mummenhoff, 2001; Koch & Al-Shehbaz, 2004). High [Zn]_shoot is probably a general feature of Noccaea and its sister clade Raparia (Macnair, 2003; Taylor, 2004), but not of Thlaspiceras, which nevertheless contains Zn-hypertolerant species (e.g. Thlaspiceras oxyceras (Boiss.) F. K. Mey; Peer et al., 2003), nor of the more distantly related nonZn-hypertolerant Microthlaspi and Neurotropis clades. Thus, the high [Zn]_shoot character most likely evolved at the base of the Noccaea/Raparia clade, or less likely at the base of the Noccaea/Raparia/Thlaspiceras clade with a subsequent reversion to the low [Zn]_shoot character in Thlaspiceras (Macnair, 2003; Taylor, 2004). Intriguingly, since Ni hyperaccumulation also evolved at the base of the Noccaea/Raparia/Thlaspiceras clade, high [Zn]_shoot may be a modification of the Ni hyperaccumulation character (Taylor, 2004). Nickel hyperaccumulation is more common than Zn hyperaccumulation among angiosperms (> 300 species, in c. 34 families), although > 80% of temperate Ni hyperaccumulators are Brassicaceae species (Reeves & Baker, 2000; Borhidi, 2001). The only other Brassicaceae species with an unequivocally high [Zn]_shoot character is Arabidopsis halleri. Thus, Zn hyperaccumulation has probably only arisen during two relatively recent evolutionary events within the Brassicaceae, and possibly on very few isolated occasions elsewhere in the angiosperms. The following sections focus primarily on Thlaspi caerulescens.

Two unusual root properties of Zn hyperaccumulators Two remarkable features of T. caerulescens roots may be linked to its ability to hyperaccumulate Zn. First, a zincophilic root foraging response to heterogeneous [Zn]_soil has been shown, analogous to the exploitation of spatially heterogeneous soil macronutrients (Schwartz et al., 1999; Whiting et al., 2000; Haines, 2002). Increases in root biomass (including root length and root hair production) occur in high-Zn-containing patches compared with adjacent Zn-deficient patches. These responses are not thought to be constitutive at the species level (Whiting et al., 2000; Haines, 2002) and warrant further genetic and molecular investigations. Second, a ‘peri-endodermal’ layer of cells with irregularly thickened inner tangential walls extending to < 1 mm from the root tip has been reported recently in T. caerulescens (Zelko et al., in press; 2, 3). This layer is composed of secondary cell walls impregnated by suberin/lignin, forming a compact cylinder surrounding the endodermis from the outer side. This layer is not seen in T. arvense (Zelko et al., in press) or A. thaliana (van de Mortel et al., 2006) when compared with T. caerulescens. The development of the endodermis also differs between T. caerulescens and T. arvense (Zelko et al., in press). In T. caerulescens, Casparian bands (the first stage of endodermal development) form < 1 mm from the root tip, and suberin lamellae (the second stage of endodermal development) are formed in all endodermal cells c. 5–6 mm from the apex. In T. arvense, Casparian bands develop c. 2 mm from the root tip and suberin lamellae are formed in all endodermal cells > 10 mm from the root tip. Although a similar feature to this peri-endodermal layer was observed by early anatomists in some Brassicaceae, being described as ‘réseau sus-endodermique’ in 1887 (van Tieghem, 1887; Zelko et al., in press), the precise function of this layer of cells and its impact on apoplastic and symplastic fluxes of Zn into the root stele (and its effects on Zn efflux from the stele) remain unclear, as does the distribution of this character among closely related taxa. Further anatomical studies in the Thlaspiceras/Noccaea/Raparia/Microthlaspi/Neurotropis clade are likely to be revealing in this respect. In a closely related Brassicaceae species, Thellungiella halophila O. E. Schulz, a second layer of endodermis is developed and it is probably related to the salt tolerance of this halophyte (Inan et al., 2004). Thus, the structural/functional adaptations of roots associated with metallophily are highly variable in this interesting group of plants.