The 19 metal(loid) ions tested showed significant differences in toxicity to lettuce roots, with {EC50} values in the order of magnitude range of 10− 2-103. Some ions, such as Ag+ and Se6+, were highly toxic to lettuce roots, with {EC50} of 0.05 µM and 0.15 µM, respectively. The highly toxic effects of Ag+ have been widely reported in aquatic and soil organisms. For example, Morgan and Wood (2004) and Le et al. (2013) tested the toxicity of Ag to rainbow trout (Oncorhynchus mykiss) and lettuce (Lactuca sativa) and found that 0.03 µM and 0.13 µM Ag could inhibit 50% of growth, respectively. High toxicity of Ag to plant roots can be attributed to its strong binding to the cell wall and inhibition of enzymes required for cell expansion to produce rupturing, etc. (Blamey et al. 2010). The toxicity of Se6+ ({EC50} = 0.15 µM) was second only to that of highly toxic Ag+ ({EC50} = 0.05 µM). Since Se can replace S in the amino acid cysteine and methionine, its toxicity in root cells may be related to alterations in protein biosynthesis, structure and function (Van Hoewyk 2013).
There are often differences in toxicity and toxicity symptoms for different valences of an element. In the present study, we observed differences in the toxic effects and symptoms of different valences of As on lettuce. The toxicity of As3+ to lettuce roots ({EC50} = 2.75 µM) was slightly higher than that of As5+ ({EC50} = 5.87 µM), which was consistent with the results obtained by SEM observation (Fig. 1b-c). However, Kopittke et al. (2012) obtained different results in research on the toxicity of different valences of As to cowpea (Vigna unguiculata) roots, finding that 3.6 µM As3+ and 0.9 µM As5+ could reduce the elongation of cowpea roots by 50%. Difference between the results of the two studies may be due to the different plants tested. Nonetheless, both studies showed differences in the toxicity of the variable-valence element As, which could be attributed to its different phytotoxicity mechanisms. Studies have shown that the phytotoxicity of As3+ [absorbed through the silicic acid transport system (Asher and Reay 1979; Ma et al. 2008)] is due to its reaction with dithiol groups on proteins and inhibition of enzyme reactions requiring free sulfhydryl groups (Horswell and Speir 2006). On the contrary, As5+ competes with phosphate and is absorbed through phosphate transporters (Asher and Reay 1979; Zhao et al. 2009). Similarly, the phytotoxicity and toxicity symptoms of different valences of Se were also quite different (Table 1, Fig. 1d-e). There is a high affinity between Se6+ and sulfate transporters, which is conducive to absorption and transport (Zhang et al. 2003). However, Se4+ is transported through the cytoplasm, and phosphate transporters are involved in this process (Hopper and Parker 1999; Li et al. 2008).
Macroelements, such as Ca and Mg, are essential nutrients for plant growth and development and play an important role in a series of physiological and biochemical reactions. These elements are generally not toxic and might even compete with toxic ions for bioactive sites, so as to reduce the toxicity of the toxic ions (Kopittke et al. 2012). However, when their concentration is high or exceeds the required concentration for plant growth, these ions will inhibit plant growth and produce toxic effects. Kopittke et al. (2011) found that approximately 14000 µM Mg2+ inhibits cowpea root elongation by 50%, suggesting that the toxicity of Mg is much lower than that of other elements such as Ag and Cu.
Compared with previous studies (Meng et al. 2019a), the QICAR equations established in the present study on the basis of relationships between toxicity and ionic characteristics of 19 metal(loid)s have a poor predictive effect on toxicity in lettuce (R2Adj = 0.238–0.503, p < 0.05). This may be due to the large variety of metal(loid)s and the large differences in physicochemical properties of elements, resulting in different dose-responses between organisms and individual elements (Luo et al. 2021; Zamil et al. 2009). Hence, classification of elements with similar properties may improve the predictive performance of the model. Soft-hard acid-base theory (HSAB) was developed on the basis of Lewis acid-base theory, which divides ions into soft and hard categories according to the properties of elements, where "soft" refers to those ions with larger radius, higher polarizability and lower charge density, while hard ions have the opposite properties (Pearson 1963, 1968; Pearson and Mawby 1967).
On the basis of QICAR, both the relationships between toxicity and ion characteristics and toxicity prediction effects were significantly improved after ions were classified following the HSAB theory. The toxicity of soft ions showed significant correlations with 7 of the 23 physicochemical properties (D, Z, IP, Z/AR, Z/r, |log(-KOH)|, Z2/r; p < 0.05); D showed the largest correlation coefficient with log{EC50}, making it the best variable for establishing the toxicity prediction equation. D is density, a physical property of metal(loid) ions. In general, periodic variation in density is determined by atomic weight, atomic volume and lattice type (Enache et al. 2003). Other properties, such as charge Z, play an important role in the interaction between soft ions and soft ligands as soft receptors usually exhibit low charge (Pearson 1963, 1968). The toxicity of hard ions showed a significant correlation with only one property (AR/AW). AR/AW is the ratio of atomic radius to atomic weight, which characterizes the electron density of metal(loid) ions. Generally, the charge density of the acceptor and donor is the main factor in the interaction between hard ions and hard ligands (Ahrland 1968). However, due to the reduction in the species and number of metal(loid) elements after grouping, higher R2 values obtained using the soft-hard ion grouping method of QICAR should be further verified using more elements.
The soft-hard ion grouping led to a reduction in the number of samples, and we could not establish a unified QICAR method. We therefore further explored the relationships between ion toxicity and the three ligand-binding constants (σCon, HLScale and log K) that characterize the binding affinity of metal(loid)s to ligands. Each parameter had different prediction effects on metal(loid) phytotoxicity. σCon showed a strong correlation with toxicity (p < 0.001), with the best fitting effect (R2Adj = 0.844), indicating that ion softness is a key factor in evaluating and predicting metal(loid) toxicity to lettuce roots and that interactions between soft ions and soft ligands play a major role in metal(loid) toxicity to lettuce. Similarly, Kinraide (2009) explored the relationship between metal(loid) toxicity and σCon and found that the interaction of charge (Z) and σCon achieved good toxicity prediction effects (R2 = 0.923); however, this was limited to low-valent elements (Z ≤ 3).
By contrast, toxicity was not significantly correlated with log K or HLScale (p > 0.05). Customarily, log K is directly proportional to EC value (Meng et al. 2019a). However, log K values of some ions such as Cu2+ and Ag+ in the present study corresponded to {EC50} in the opposite relationship (Table 1). A possible reason for this trend was that log K values were obtained from previously published literature, while {EC50} values were derived from our experiments; the lettuce varieties used in the studies of these two types of data may be different, thereby resulting in a non-corresponding relationship between log K and toxicity for some elements. However, when Cu2+ was removed, log K was significantly correlated with toxicity and showed a better toxicity prediction effect (R2Adj = 0.767, p = 0.033). In addition, the log K method currently available in the published literature involves limited metal species and toxicity data. There are few studies on the evaluation and prediction of lettuce toxicity using BLM. We could only obtain log K values of six metal elements (Table 1). Therefore, obtaining the log K of more elements will be more meaningful for modeling.
A poor correlation was shown between HLScale and toxicity, which may be caused by differences in metal(loid) phytotoxicity mechanism. Metal cations binding to hard ligands, regarded as a common mechanism, results in direct toxicity through inhibiting the controlled loosening of cell walls (Kopittke et al. 2011). Nevertheless, this mechanism may not work for all ions; that is to say, the binding strength of the ion to the hard ligand may not be dominant in all ion-induced toxicity. For example, the soft ion Ag+ tends to bind strongly to RS- functional groups (soft ligands) in metallothionein to exert toxic effects (Bell et al. 2002), independent of the binding strength of hard ions (Kinraide 2009). Nevertheless, whether this applies to more plant species or elements requires further study.