Environmental Pollution 202 (2015) 41e49

Contents lists available at ScienceDirect

Environmental Pollution journal homepage: www.elsevier.com/locate/envpol

Development and validation of a terrestrial biotic ligand model for Ni toxicity to barley root elongation for non-calcareous soils Yanqing Lin, Dominic M. Di Toro, Herbert E. Allen* Center for the Study of Metals in the Environment, Department of Civil and Environmental Engineering, University of Delaware, Newark, DE 19716, USA

a r t i c l e i n f o

a b s t r a c t

Article history: Received 24 November 2014 Received in revised form 10 March 2015 Accepted 12 March 2015 Available online 21 March 2015

A Terrestrial Biotic Ligand Model (TBLM) for Ni toxicity to barley root elongation (RE) developed from experiments conducted in sand culture was used to predict toxicity in non-calcareous soils. Ca2þ and Mg2þ concentrations and pH in sand solution were varied individually and TBLM parameters were computed. EC50 increased as Mg2þ increased, whereas the effect of Ca2þ was insignificant. TBLM parameters developed from sand culture were validated by toxicity tests in eight Ni-amended, noncalcareous soils. Additional to Ni2þ toxicity, toxicity from all solution ions was modelled independently as an osmotic effect and needed to be included for soil culture results. The EC50s and EC10s in soil culture were predicted within twofold of measured results. These are close to the results obtained using parameters estimated from the soil culture data itself. © 2015 Elsevier Ltd. All rights reserved.

Keywords: Barley Root elongation Sand culture Terrestrial biotic ligand model Osmotic effect Soil

1. Introduction Understanding nickel toxicity in soils is necessary to properly address environmental quality criteria and standards (Janssen et al., 2000). Total metal concentration does not account for metal bioavailability and toxicity to the soil ecosystem (Allen, 2002). Metal bioavailability and toxicity in aquatic and sediment systems have been predicted based on its speciation and competition from other cations in solution (Di Toro et al., 2001, 2005). The terrestrial biotic ligand model (TBLM) similarly addressed soil (Steenbergen et al., 2005; Thakali et al., 2006), in which partitioning of nickel and its toxicity in the soil system are considered (Supplementary data, Fig. SD1). Several studies have been conducted to develop the TBLM for Ni toxicity in controlled hydroponic water solution culture systems (Antunes and Kreager, 2009; Li et al., 2009; Lock et al., 2007c). However, these were not used to model Ni toxicity in soil systems for validation. Lock et al. (2006; 2007b) were able to predict Co toxicity to potworm survival in an artificial soil and a field soil within a factor of 2, but only within a factor of 4 for Ni toxicity to barley root elongation. Wang et al. (2011b) were able to predict

* Corresponding author. E-mail address: [email protected] (H.E. Allen). http://dx.doi.org/10.1016/j.envpol.2015.03.015 0269-7491/© 2015 Elsevier Ltd. All rights reserved.

EC50 for Ni2þ toxicity to barley root elongation for non-calcareous soils within a factor of 2 based on an electrostatic toxicity model. Thakali et al. (2006) predicted Ni toxicity to barley root elongation (RE) in NiCl2 or CuCl2 amended soils, using TBLM parameters developed from the soil culture itself. Adding Ni2þ to soil causes release of protons and major cations to the soil solution (Ponizovsky et al., 2006b). Therefore, it is difficult to control soil solution composition to study the effect of cations individually and independently. Constant composition can be achieved in hydroponic systems with water only or with inert sand, but the solution (hydroponic sand solution or hydroponic water solution, which we refer to as sand solution or water solution) must be frequently renewed as these systems lack soil's buffering capacity. This study was conducted to develop TBLM for Ni toxicity in an inert sand culture system, in which the cations and pH were varied individually and independently, and to use those values to predict the Ni toxicity in non-calcareous soil cultures independent of the systems used to develop the model's constants. The soil toxicity data used for validation was taken from previous work (Rooney et al., 2007; Thakali et al., 2006). Competitive effects of Hþ, Ca2þ, and Mg2þ on Ni toxicity to barley root elongation in the sand culture were studied individually. Prediction of soil culture data by the parameters obtained from the sand system was compared with the prediction using model parameters estimated from soil culture data itself. The present study also provided a basis for additional work

42

Y. Lin et al. / Environmental Pollution 202 (2015) 41e49

(Lin, 2014) including comparison of the predictions obtained by including the activity at the root plasma membrane surface (Kinraide, 2006) and by using TBLM parameters obtained from water solution culture.

The eight soils had a wide range of properties, with soil solution pH from 3.5 to 7.2, and organic carbon content from 1.1% to 33.1%. The soils were not leached with water after NiCl2 had been added and prior to testing although Ni toxicity would have been reduced if the Ni-amended soil had been leached (Li et al., 2011).

2. Materials and methods 2.4. TBLM equations 2.1. Toxicity bioassays Toxicity testing of barley, Hordeum vulgare (Doyce), was conducted according to the ISO 11269-1 (ISO, 1993) method for the measurement of inhibition of root growth. Barley plants were grown for four days in acid washed quartz sand that containing defined composition solution. Then root length was determined, and % root elongation (%RE, root elongation ¼ root length after harvest  root length before planting) relative to the control bioassay was calculated for each plant harvested (see Supplementary data for experimental procedures). Each experiment contained 10 bioassays, including 8 exposures which, based on parameters from Thakali et al. (2006), were predicted to result in 5%, 10%, 25%, 50%, 75%, 90%, 95%, and 100% of RE relative to control, and 2 additional system controls (discussed in Supplementary data). Each bioassay had 12 barley plants, except for the exposure control run with no added Ni2þ (%RE ¼ 100) which had 24 plants. Barley RE was measured after harvest (on day 4) and the average and standard deviation were calculated. RE values that exceeded 2 standard deviations of average root elongation were excluded. On average, 5.6 of the 132 measurements were excluded from each of the total 15 experiments. 2.2. Physical and chemical measurements To compensate for evaporative water loss in sand culture the solutions were renewed daily. An Orion 370 pH meter with an accupHast combination pH electrode was used to determine pH. Ni, Ca, and Mg were measured using a PerkinElmer Optima 5300 ICP e OES. Measured pH was within 0.2 units and measured metal concentrations were within 15% of nominal values. 2.3. Soils for validation The TBLM developed in this study was validated by prediction of RE for barley grown in the soils reported by Thakali et al. (2006) and Rooney et al. (2007). These soils were amended with different amounts of NiCl2 required to bracket their total metal EC50 concentration (Oorts et al., 2006). Following Thakali et al. (2006), the eight non-calcareous soils having OC content > 1% were selected because of the unavailability of a speciation model able to accurately predict nickel partitioning for calcareous soils or those with low soil organic matter (SOM, characterized by OC). Activities of Ni2þ and other cations in soil solution for the eight soils were predicted with the soil properties using the WHAM computation (Thakali et al., 2006), assuming that Ni was mainly bound to SOM. For high pH soils metal precipitation with CO2 3 must be considered, and for low OM soils (OC < 1%) it is likely that Ni is not bound only to SOM. These factors will not allow prediction of Ni partitioning for the calcareous or low OM soils (Ponizovsky et al., 2006a). The Ni concentrations of the low SOM soils for which OC < 1% cannot be predicted within a half log unit using WHAM VI, whereas the selected soils were predicted with that accuracy (Ponizovsky et al., 2008). The other reason for choosing non-calcareous soils was that soils with low pH are more prone to exhibiting metal toxicity than are soils with higher pH. Metal toxicity is generally reduced after soil pH increases (Lexmond, 1980; Smith, 1994).

The TBLM assumes that toxicity is related to Ni2þ binding with the biotic ligand (BL) sites, and competition of Hþ, Ca2þ, and Mg2þ for the BL alleviates toxicity. Toxicity is determined by the fraction (f) of total BL bound by Ni2þ (Di Toro et al., 2001; Thakali et al., 2006):

f¼ ¼

½NiBL ½TBL


 KNiBL Ni2þ



 1 þ KNiBL Ni2þ þ KHBL Hþ þ KCaBL Ca2þ þ KMgBL Mg2þ (1)

where [NiBL] is the concentration (moles g1) of BL sites bound by Ni, [TBL] is total BL sites (moles g1), and KXiBL (L mol1) is the conditional binding constant for specific cationeBL interaction for an endpoint (representing the affinity of cation Xi for BL), and {} represents cation activity. þ It was found that NiHCOþ 3 was toxic (Li et al., 2009), and NiOH was possibly toxic as well (Antunes et al., 2012; De Schamphelaere and Janssen, 2002). In our system with inert sand, the only carbonate source was from the atmosphere, so {NiHCOþ 3 } was low compared to {Ni2þ}. Using MINEQLþ (Environmental Research Software, Hallowell, ME) (Schecher and McAvoy, 1992) we found the {Ni2þ} is more than 105 times that of {NiHCOþ 3 } for pH 4 and about 140 times for pH 7. For our sand solutions at pH 4e7 the {NiOHþ} was even lower and was correlated to {NiHCOþ 3 }, with the {NiOHþ} about 21% of the {NiHCOþ 3 }, which was similar to the results of Li et al. (2009) in hydroponic water solution. If we consider the toxicity from the predominant free Ni2þ only, the toxic response that correlates to the fraction of the total BL bound by the free Ni2þ using the logelogistic dose response function is:

R¼

100  b 1 þ ff 50

100

¼ 1þ

2þ

KNiBL fNi g 2þ f50 ð1þKNiBL fNi gþKHBL fHþ gþKCaBL fCa2þ gþKMgBL fMg2þ gÞ

!b

(2)

where R ¼ biological response as % of the control and for the control bioassay R ¼ 100, f50 is the fraction of the total BL sites occupied by Ni2þ for a 50% response, and b is the doseeresponse shape parameter. The EC50 is derived from Eq. (2) with R ¼ 50%:

o n EC50 Ni2þ ¼

 n o f50 1 þ KHBL Hþ 50 ð1  f50 ÞKNiBL n o n o  2þ þ KCaBL Ca þ KMgBL Mg2þ 50

50

(3)

where the subscript “50” represents the activity of a cation at the 50% effect level. Eq. (2) can be used to calculate the toxic response as percent of the control, and Eq. (3) can be used to calculate the EC50. If the toxicity of NiHCOþ 3 is also considered, Eq. (2) can be modified to include binding of NiHCOþ 3 to the BL:

Y. Lin et al. / Environmental Pollution 202 (2015) 41e49

R¼

1þ@ ¼

100

0 KNiBL fNi f50 ð1þKNiBL fNi

2þ

2þ

B B 1þB B @

1b

gþKNiHCO3 BL fNiHCO3 g þ

gþKNiHCO3 BL fNiHCO3 gþKHBL fHþ gþKCaBL fCa2þ gþKMgBL fMg2þ gÞ þ

A

100

0

KNiBL þKNiHCO3 BL

f50

1þ

KNiBL þKNiHCO3 BL

! fNiHCOþ3 g fNi2þ g

1b

(4)

C C !C C A

fNi2þ g

! fNiHCOþ3 g fNi2þ gþK fNi2þ g

43

2þ þ gþKMgBL fMg2þ g HBL fH gþKCaBL fCa

and similarly to Eq. (3), the EC50 including binding of NiHCOþ 3 is:


2þ 


o f 1 þ K
Hþ  þ K n þ KMgBL Mg2þ 50 HBL 50 CaBL Ca 50 50 ! EC50 Ni2þ ¼ fNiHCOþ3 g50 ð1  f50 Þ KNiBL þ KNiHCO3 BL 2þ fNi

g50






f50 1 þ KHBL Hþ 50 þ KCaBL Ca2þ 50 þ KMgBL Mg2þ 50 n o   ¼ ð1  f50 Þ KNiBL þ KNiHCO3 BL KNiHCOþ HCO 3 3

Toxicity of NiOHþ, in addition to that of Ni2þ and NiHCOþ 3 , can likewise be included in the equations for metal binding to the BL and for toxicity as shown in Supplementary data.

2.5. Data treatment and statistics Chemical speciation in the sand culture solution was calculated using MINEQLþ (Schecher and McAvoy, 1992). Chemical speciation in soil and soil solution was calculated using WHAM VI (The Windermere Humic Aqueous Model, Version 6.0.10, Natural Environment Research Council, 2001) (Tipping, 1998). For these calculations, the stability constant for NiCl02 formation from the MINEQL þ database (Ni2þ þ 2Cl ¼ NiCl02; log K ¼ 0.96) was incorporated into the WHAM database. SOM was estimated as 1.8 times the soil organic carbon (OC) content (Sparks, 2003). SOM was assumed to be 84% humic acid and 16% fulvic acid, and dissolved OC is 65% active as colloidal fulvic acid (Tipping et al., 2003). CO2 partial pressure (pCO2) was set to 3.16  104 atm and the only source of carbonate for experiments was from the atmosphere. Linear relationships between released Ca2þ, Mg2þ, Kþ, and Naþ and adsorbed Ni cationic species (Ponizovsky et al., 2006b) were estimated from soil solution data, and these calculated cation concentrations were used to compute the molarity of ions in soil solution. Activity coefficients were calculated using the DebyeeHückel equation which was used to convert concentrations of chemical species to activity. The observed EC50 {Ni2þ} for 4-day barley root elongation was estimated from the doseeresponse relationship between {Ni2þ} and average RE using TRAP (Toxicity Relationship Analysis Program, version 1.0, (EPA, 2002)), with the logistic regression equation for each experiment data set:

(5)

50

RE ¼

RE0 RE0 ¼ 


0 2þ 1 þ ebðlogfNi glog EC50Þ 1 þ Ni2þ EC50 b

(6)

where RE0 is the root elongation at the control condition and b and b0 (b0 ¼ b/Ln10) are the slope parameters. The value of RE0 was used as the basis for calculating the %RE response. The observed EC50 total Ni for soil bioassays were calculated by TRAP in the same way. The EC10 value for soil bioassays was estimated similarly using Eq. (6) with the estimated EC50 and slope parameters. The BL constants were calculated by the linear relationships between EC50 free Ni2þ activity and the activities of Hþ, Ca2þ and Mg2þ at the 50% effect level (Eq. (3)) using a multilinear regression method, a modification of the approach by De Schamphelaere and Janssen (2002). The root mean square error (RMSE) of prediction was minimized using Solver in Microsoft Excel 2003. Standard errors of TBLM parameters were estimated using SOLVER AID (De Levie, 2004). 3. Results and discussion 3.1. Estimation of TBLM parameters using sand culture data A linear relationship between EC50 activities of Hþ, Ca2þ and Mg2þ at the EC50 was predicted by Eq. (3) and results are in Fig. 1aec. The hollow squares with dashed line are baseline values about which variations in activities are made, and represent a constant EC50 if the variations in pH, {Ca2＋}or {Mg2þ}have no effect on toxicity. Fig. 1c shows that EC50 increases linearly from 0.14 to 0.48 mM as {Mg2þ}increases from 0.6 to 7.8 mM (R2 ¼ 0.91; p < 0.05), suggesting that competition by Mg2þ alleviates Ni toxicity. EC50 activity variation (0.09e0.17 mM) for {Ca2þ} in excess of baseline activity (baseline value 0.14 mM) was smaller than that for

44

Y. Lin et al. / Environmental Pollution 202 (2015) 41e49

2þ 2þ Fig. 1. Relationship between EC50 for the Ni2þ activity and (a) pH with consideration of toxicity of NiHCOþ for the sand culture test. Symbol , 3 , the activity of (b) Ca , and (c) Mg represents the constant values of pH, Ca2þ, or Mg2þ used when each of the other values are varied individually The dashed line represents a constant EC50 relative to the constant values.

{Hþ} and (Mg2þ} and not linear (Fig. 1b). Toxicity was enhanced initially with increasing {Ca2þ}, perhaps due to an effect on the plasma membrane (Wang et al., 2010, 2011a) and then remained constant. In this work no statistically significant binding constant by Ca2þ was found, indicating that the EC50 competition by Ca2þ (Table 1, Fig. 1b) was insignificant with {Ca2þ} from 0.19 to 27 mM. He et al. (2014) also indicated the effect of {Ca2þ} on Ni uptake and toxicity differed from {Hþ} and {Mg2þ}; Ca2þ may directly affect physiological functions dependent on membrane permeability, not simply explained by competition. The competition for Ni2þ binding by Hþ for pH higher than 5 is reduced and the toxicity is increased relative to that at lower pH. Similar results were observed in Ni, Cu and Co toxicity tests to barley root elongation in water culture (Lock et al., 2007a, 2007b;

2007c). It is possible that Ni species present at high pH, e.g. NiHCOþ 3 , results in enhanced toxicity (Li et al., 2009). The potential toxicity from NiHCOþ 3 was considered. The binding constants KHBL, KCaBL, KMgBL and KNiHCO3 BL were estimated using Eqs. (3) and (5) with the multiple linear regression method by minimizing the RMSE of EC50. The value of f50, b and KNiBL were estimated using Eq. (2) by minimizing the RMSE of %RE prediction (Table 1). Determination of a unique estimate for f50 is not possible because error in the estimates of f50 and KNiBL are large. In our study estimated f50 and KNiBL are highly correlated (R2 ¼ 0.96 for a linear regression between selected f50 from 0.05 to 0.6 and the respective estimated log KNiBL values, p < 0.05). The corresponding prediction RMSE of %RE is 9.4% for f50 ¼ 0.4e0.6, slightly increasing to 10.0% for f50 ¼ 0.05. This is also seen in the residual error surface

Table 1 TBLM parameters (± Standard Errora) for barley root elongation estimated from bioassays in sand cultures of current study, and from previous works in soil and water culture. Growth culture

Hydroponic sand

Soil

Hydroponic water

Hydroponic water

Hydroponic water

f50

0.4 3.89 ± 1.41 3.70 ± 0.39 4.45 ± 0.77 3.39 ± 0.56 ec 2.49 ± 0.051 5.67 ± 0.13 0.160 ± 0.012 1.92 ± 0.25 Current study

0.05 2.37 ± 0.33 3.60 ± 0.53 eb 4.52 ± 0.62 1.50 3.81 ± 0.60 eb eb eb Thakali et al. (2006)

0.57 eb 5.27 eb eb eb 3.77 eb eb eb Lock et al. (2007c)

0.36 ± 0.08 2.67 ± 0.39 4.83 ± 0.17 5.36 ± 0.17 4.29 ± 0.12 1.60 ± 0.15 4.01 ± 0.03 eb eb eb Li et al. (2009)

eb eb 3.1 eb 5.4 3.3 4.6 eb eb eb Antunes and Kreager (2009)

b

log KNiBL log KNiHCO3 BL log KHBL log KCaBL log KMgBL Osmotic RE0, cm Osmotic M50, M Osmotic b2 Source a b c

Estimated using SOLVER AID (De Levie, 2004). Not reported. No significant binding is estimated.

Y. Lin et al. / Environmental Pollution 202 (2015) 41e49

Fig. 2. Predicted vs. observed root elongation (as % of the control) for sand culture by TBLM parameters developed from sand culture data and Eq. (2). The solid line represents the 1:1 ratio and the dash lines represent a factor of 2 variation above and below the 1:1 line.

for estimations from soil tests (Thakali et al., 2006). Considering previous reported values of f50 ¼ 0.57 by Lock et al. (2007c) and f50 ¼ 0.36 by Li et al. (2009), and the range of our estimated f50, the value f50 ¼ 0.4 was selected. Using values of f50 and b from Table 1, the toxic response in %RE from the sand culture experiments was predicted (R2 ¼ 0.90, p < 0.001) within a factor of 2 with RMSE of 9.4% (Fig. 2). Including toxicity of NiHCOþ 3 slightly improved the prediction of sand culture data. A direct comparison between observed and estimated (solid line in Fig. 1a) EC50 including toxicity of NiHCOþ 3 based on Eq. (5) for different pH was conducted to study the effect þ of NiHCOþ 3 . A binding constant between NiHCO3 and biotic ligand 2þ 3.39 times of that of Ni with the biotic ligand, suggested by Li et al. (2009), reduced the sand culture RMSE %RE from 9.40% to 9.39%, and the lowest RMSE prediction resulted from including þ toxicity of NiHCOþ 3 as binding constant between NiHCO3 and BL as 5.7 times of that of Ni2þ, with RMSE of 9.38%. The toxicity of NiHCOþ 3 in soil culture prediction was studied as well (details are below), and considering toxicity of NiHCOþ 3 with binding constant to BL as 3.39 and 5.7 times that of Ni2þ slightly increases the RMSE %RE by 0.07% and 0.13% respectively. The best RMSE prediction resulted from including NiOHþ toxicity with a binding constant between NiOHþ and biotic ligand

Fig. 3. Doseeresponse between root elongation and the molarity produced by Hþ, Ca2þ, Mg2þ, or Naþ salts. The root elongations were normalized to the sand system control results.

45

26.7 times of that of Ni2þ (Supplementary data), which only reduced the sand culture RMSE %RE from 9.40% when only Ni2þwas considered, to 9.39%. It is not necessary to describe the NiOHþ toxicity separately because {NiOHþ} is less than {NiHCOþ 3 } and {NiOHþ} is correlated to {NiHCOþ 3 }. Overall, including toxicity of 2þ NiHCOþ gave an improved prediction, 3 as 5.7 times of that of Ni but those improvements are small as the concentration of NiHCOþ 3 is very low in our inert sand with hydroponic solution that does not contain carbonate other than that from adsorption from the atmosphere and for non-calcareous soil systems. Considering the þ toxicity of free Ni2þ only, without NiHCOþ 3 and NiOH , will result in a similar prediction (RMSE ¼ 9.40%) compared to 9.38% for Ni2þ and Ni complexes. Similarly Antunes et al. (2012) indicated that Cu carbonate complexes were bioavailable for plant roots but the BLM framework without considering the toxicity of Cu carbonate complexes achieved better prediction. Additional information including different toxicity effects of þ þ NiHCOþ 3 and NiOH , and assuming a separate H binding constant for pH < 5 and pH > 5, are addressed in Supplementary data. The Ni TBLM parameters for barley root elongation from previous works are presented in Table 1. The value of log KNiBL ¼ 3.70 for the current study is close to the value of 3.60 reported by Thakali et al. (2006) and 3.10 reported by Antunes and Kreager (2009). TBLM parameters from water culture (Li et al., 2009; Lock et al., 2007c) have greater conditional binding constant for Ni2þ and competing cations, except for one KNiBL that was developed for the low affinity ligands (Antunes and Kreager, 2009). It appears that the cations' binding to the root biotic ligands were enhanced in water only exposure compared to those in sand culture (Allen et al., 2008). Villagarcia et al. (2001) reported that plants grown in water only culture are more sensitive to Al than are plants grown in sand culture.

3.2. Osmotic effect In control exposures with no Ni2þ added, the root elongation decreased as Ca2þ or Mg2þ in solutions in the sand cultures increased. This was interpreted as an osmotic effect, which is related to the total molarity of salinity, increasing the potential required by the plant to take up water (Ferguson and Grattan, 2005; Kinraide, 1999; Kopittke et al., 2011; Yilmaz, 2007). The osmotic effect is not caused by toxic Ni species as the effect is present in solutions in which Ni is absent. Therefore, it is different than the influence that the plasma membrane surface potential exerts by modifying the activity of a toxic metal like Ni2þ at the interface of the plant root and the solution. However, the ions producing the osmotic effect such as Ca2þ or Mg2þ also affect the plasma membrane surface potential as shown by Kinraide (2006) to influence the toxicity of nickel and other ions. The cations in the solution will not only affect metal toxicity as described in the TBLM (Eq. (2)) or by the plasma membrane theory, but they also produce additional toxicity that is independent of Ni2þ toxicity. A doseeresponse curve between root elongation and molarity is shown in Fig. 3 using the data for which only Ca2þ and Mg2þ were added and pH was adjusted but no Ni was added. The root elongations were normalized using the sand system control. The Naþ series data came from an additional experiment (Supplementary data section 3.3) to study the effect of added NaCl on barley root elongation. Fig. 3 shows that increasing molarity of these cations in the absence of Ni reduces barley root elongation, and the dose response follows a logelogistic relationship that is independent of which cation has contributed to the molarity. To include the osmotic effect, the simplest formulation is to assume a product dependency (Wang et al., 2013; Yermiyahu et al., 1997):

46

Y. Lin et al. / Environmental Pollution 202 (2015) 41e49

Fig. 4. Prediction of barley root elongation, total Ni EC50 (C) and EC10 (◊) in soil culture. (a). Predicted vs. observed root elongation (as % of the control) for soil culture using sand TBLM parameters developed from sand culture including osmotic effect. (b). Predicted vs. observed root elongation for soil culture using sand parameters without including osmotic effect. (c). Predicted vs. observed root elongation for soil culture using soil parameters developed from soil culture itself. (d). Predicted vs. observed total Ni EC50 and EC10 for soil culture using sand parameters including osmotic effect toxicity. (e). Predicted vs. observed total Ni EC50 and EC10 for soil culture using sand parameters without including osmotic effect toxicity. (f). Predicted vs. observed total Ni EC50 and EC10 for soil culture using soil parameters developed from soil culture itself. The solid line represents the 1:1 ratio and the dash lines represent a factor of 2 variation above and below the 1:1 line. Eq. (6) was used for prediction with osmotic effect, and Eqs. (2) and (3) were used for prediction without osmotic effect.

100

R¼ 1þ

2þ

KNiBL fNi g 2þ f50 ð1þKNiBL fNi gþKHBL fHþ gþKCaBL fCa2þ gþKMgBL fMg2þ gÞ

 1þ

1  b2 M M50

!b1

(7)

where M50 is the molarity corresponding to 50% inhibition of root elongation, b2 is a shape parameter required to account for the osmotic effect. The M50 (0.16 ± 0.012 M) and b2 (1.923 ± 0.249) were estimated from the logelogistic response curve in Fig. 3. As the osmotic effect results from total molarity in the solution, not specific ions, and the osmotic effect from different cations (Hþ, Ca2þ, Mg2þ, and Naþ) as shown in Fig. 3 follow the same trend (R2 ¼ 0.98, p < 0.001), so the osmotic effect parameters developed here are applicable to other metal toxicity systems as well as for

Y. Lin et al. / Environmental Pollution 202 (2015) 41e49

47

Table 2 Comparison of total Ni EC50, EC10 and RMSE %RE for soil culture toxicity prediction using the TBLM parameters developed from sand culture (current study) including the osmotic effect, without including the osmotic effect, and TBLM parameters that were developed from soil culture itself from previous work. Growth culture

Sand W/osmotic effect Sand W/O osmotic effect Soil (Thakali et al., 2006) a b

EC50

EC10

RMSE %RE

Average ratioa

Average differenceb

Average ratio

Average difference

1.15 1.28 1.03

0.020 0.080 0.0038

1.12 1.37 1.17

0.015 0.099 0.034

14.1 15.2 10.5

Average value of (predicted EC50/observed EC50) or for EC10 for each soil tested. Average value of (log Predicted EC50  log Observed EC50) or for EC10 for each soil tested.

nickel. Osmotic toxicity must be considered in the soil experiments. Ni dosage is large because most added Ni is sorbed to the soil particles and other cations are released to the soil solution (Ponizovsky et al., 2006b). Ni toxicity was overpredicted and the EC50 value was underestimated (Table SD2) if osmotic toxicity was excluded. Although increasing Mg2þ will decrease Ni2þ toxicity (via ion competition reactions), it also produces osmotic toxicity. These two effects must be described separately to account for the net effect on toxicity. In sand solution there is negligible sorption of Ni, so the molarity change caused by the addition of Ni is very small and can be considered insignificant. Although accounting for the osmotic effect does not significantly change predictions for soil exposure, it is useful to have a quantitative expression for this effect. 3.3. Predicting root elongation and EC50 for soils TBLM parameters developed from sand culture (Table 1) were used to predict root elongation, EC50 and EC10 values (Fig. 4) for soil bioassays. If the osmotic effect is not included, the %RE, total Ni EC50 and EC10 are slightly over predicted (Fig. 4b and e). Including the osmotic effect slightly improved the prediction: the RMSE of predicted %RE decreased from 15.2% (Fig. 4b; R2 ¼ 0.86, p < 0.001) to 14.1% (Fig. 4a; R2 ¼ 0.88, p < 0.001), the average ratio of predicted to observed EC50 decreased from 1.28 (Fig. 4e) to 1.15 (Fig. 4d), and the average difference between predicted and observed EC50 and EC10 also decreased (Table 2). Total Ni EC50 and EC10 were generally predicted (R2 ¼ 0.89, p < 0.001 for EC50; R2 ¼ 0.92, p < 0.001 for EC10) within a factor of 2 when the osmotic effect was accounted for in the prediction. Other than the osmotic effect, it was proposed that the adverse effect from Ca2þ deficiency, and the toxicities from specific ions such as Naþ and Cl could be important for plant response (Kopittke et al., 2011; Munns, 2002). Calcium is an essential nutrient for plant growth and cell plasma membrane integrity (Del Amor and Marcelis, 2003; Kopittke et al., 2014). A critical value of {Ca2þ}0 ¼ 1.6 mM as activity at the plasma membrane surface was reported corresponding to the RE reduction due to Ca2þ deficiency, and high loading of cations in bulk solution as {Mg2þ} 4.2 mM or with pH < 3.5 could reduce the Ca2þ activity at the root plasma membrane surface inducing Ca2þ deficiency (Kopittke et al., 2011, 2014). In our study, all pHs in the bulk solution from sand and soil bioassays are greater than 3.5, and 85% of calculated {Ca2þ}0 at plasma membrane surface are higher than 1.6 mM. For the remaining 15% of bioassays for which {Ca2þ}0 is less than 1.6 mM, the average {Ca2þ}0 was about 0.9 mM, so it is possible that root growth was affected by Ca2þ deficiency in this study. It was suggested that the Cl anion may influence the plant response to Ni toxicity, as the NiCln complex could be toxic to plants (Gopalapillai et al., 2013). To ascertain the effect of Cl on Ni toxicity, an additional experiment was performed to study the toxicity of NiCl02. However, no toxicity of NiCl02 was observed (Fig. SD3) with solution concentrations of Cl as high as 0.8 M and concentration of NiCl02 as

high as 0.6 mM (Supplementary data). Tavakkoli et al. (2010) reported that it is difficult to study specific ion toxicities independently as these effects overlap the osmotic effect, and high concentration of Cl had similar damage as higher concentration of Naþ for plant growth under salt stress. Hochman et al. (2007) also indicated that modelling suggested that the adverse effects from Naþ and Cl resulted from osmotic effects. Our study also shows that for the tests with no Ni toxicity, the effects from different cations (Hþ, Ca2þ, Mg2þ, and Naþ) and their counteranions (primarily Cl and SO2 4 ) followed the same trend and therefore were modelled as an osmotic effect as shown in Fig. 3. Considering the toxicities from Ni and osmolality only can predict % RE in sand culture with RMSE of 9.4% and in soil culture with RMSE of 14.1%, our results suggest that specific ion toxicities were not observed or not important in the study. Residual analysis was performed by plotting the difference between predicted %RE and observed %RE for sand and soil bioassays versus observed %RE, pH, Ca2þ, Mg2þ, Ni2þ, molarity, and concentration of NiCl02 (Fig. SD4). No additional relationships were observed from the residual analysis. 3.4. Toxicity prediction for soils at high pH The calcareous soils with pH > 7.2 were excluded as in the study by Thakali et al. (2006) because the speciation model WHAM VI (Tipping, 1998) does not account for metal precipitation with CO2 3 . However, consideration of precipitation likely will not allow prediction of Ni partitioning (Ponizovsky et al., 2006a). This inability of predicting Ni toxicity at high pH in soil is less important than at lower pHs since the metal toxicity to plants decreases as pH of the soil increases (Lexmond, 1980; Smith, 1994). The bioavailability of toxic metal in soil decreases as pH increases, because the metal binding to soil organic matter is stronger at high pH (low Hþ activity competition) than metal binding to the biotic ligand (Plette et al., 1999; Weng et al., 2003, 2004). Although toxicity of Ni is increased at higher pH through decrease of the Hþ competition and the formation of NiHCOþ 3 (Li et al., 2009), soils are frequently made alkaline to reduce the toxicity of Ni and other metals. For example, Siebielec et al. (2007) adjusted the pH of Nicontaminated soils by the addition of nitric acid or limestone. Highly contaminated soils were phytotoxic when acidified, but toxicity was alleviated by raising the pH of all soils except for one containing over 1.1% Ni. Thus metal toxicity for low pH soils is more important than for high pH soils, and enhancement of toxicity by species such as NiHCOþ 3 is less important than the reduction in toxicity resulting from stronger Ni binding to soil organic matter at high pH or other mechanisms that reduce Ni bioavailability. 3.5. Comparison of predicted toxicity in soil with soil and sand culture parameters Prediction of root elongation, and total Ni EC50 and EC10 for soil culture using parameters from sand culture (Table 1) including the

48

Y. Lin et al. / Environmental Pollution 202 (2015) 41e49

osmotic effect (Fig. 4a and d), were compared to those using soil parameters (Thakali et al., 2006) developed from soil culture itself (Fig. 4c and f). Predictions of root elongation in soil culture using parameters from sand culture got an RMSE %RE of 14.1%, which is close to (R2 ¼ 0.95, p < 0.001) RMSE %RE of 10.5% using soil parameters developed from soil culture data itself. The average ratio and the difference between predicted and observed EC50 using sand parameters are 1.15 and 0.020 respectively, which are close to (R2 ¼ 0.98, p < 0.001) 1.03 and 0.0038 using soil parameters (Table 2). Prediction of soil culture EC10 using sand parameters are slightly improved from using soil parameters developed from soil culture itself, with the average ratio of 1.12 using sand parameters and 1.17 using soil parameters, and the average difference of 0.015 using sand parameters and 0.034 using soil parameters (Table 2). A portion of the reason that using sand culture data was able to achieve results close to those for the soil data was the ability of the sand culture data to account for the higher osmotic effect at EC10 that was not studied separately for the soil parameters (Thakali et al., 2006). Parameters developed from sand culture and soil culture provided similar prediction for the soil toxicity data. These results suggest that TBLM parameters for soils can be derived from sand culture experiments. 4. Conclusions TBLM parameters developed from data for plants grown in sand culture were able to predict the Ni toxicity in soil culture. The competition effect from Mg2þ was more important than that from Hþ and Ca2þ. Ni toxicity in sand culture was predicted with RMSE of 9.4% using the TBLM method. The osmotic effect needs to be included in the soil culture toxicity prediction as large amounts of cations will be released as Ni salt is added to soil. Osmotic toxicity follows a logelogistic relationship and is independent of Ni toxicity. The Ni toxicity in soil culture can be predicted with RMSE of 14.1% using TBLM parameters developed from sand culture data. This is close to the RMSE of 10.5% using parameters developed from the soil culture data. Prediction of soil culture EC50 and EC10 using sand parameters are equivalent to those using soil parameters. Therefore, sand culture exposures can be used to develop the toxicity parameters applicable to soil exposures. Acknowledgements The Center for the Study of Metals in the Environment, University of Delaware funded this investigation. The Crop & Soil Environmental Sciences Department in Virginia Tech supplied barley seeds. We thank Drs. Rufus L. Chaney, Ronald T. Checkai, Michael J. McLaughlin, and Alexander Ponizovsky for their suggestions. Appendix A. Supplementary data Supplementary data related to this article can be found at http:// dx.doi.org/10.1016/j.envpol.2015.03.015. References Allen, H.E., 2002. Terrestrial ecosystems: an overview. In: Allen, H.E. (Ed.), Bioavailability of Metals in Terrestrial Ecosystems: Importance of Partitioning for Bioavailability to Invertebrates, Microbes and Plants. SETAC Press, Pensacola, FL, USA, pp. 1e5. Allen, H.E., Lin, Y.Q., Di Toro, D.M., 2008. Ecotoxicity of Ni in soil. Mineral. Mag. 72, 367e371. Antunes, P.M.C., Kreager, N.J., 2009. Development of the terrestrial biotic ligand model for predicting nickel toxicity to barley (Hordeum vulgare): ion effects at low pH. Environ. Toxicol. Chem. 28, 1704e1710.

Antunes, P.M.C., Scornaienchi, M.L., Roshon, H.D., 2012. Copper toxicity to Lemna minor modelled using humic acid as a surrogate for the plant root. Chemosphere 88, 389e394. De Levie, R., 2004. Advanced Excel for Scientific Data Analysis. Oxford University Press, New York. De Schamphelaere, K.A.C., Janssen, C.R., 2002. A biotic ligand model predicting acute copper toxicity for Daphnia magna: the effects of calcium, magnesium, sodium, potassium, and pH. Environ. Sci. Technol. 36, 48e54. Del Amor, F.M., Marcelis, L.F.M., 2003. Regulation of nutrient uptake, water uptake and growth under calcium starvation and recovery. J. Hortic. Sci. Biotechnol. 78, 343e349. Di Toro, D.M., Allen, H.E., Bergman, H.L., Meyer, J.S., Paquin, P.R., Santore, R.C., 2001. Biotic ligand model of the acute toxicity of metals. 1. Technical basis. Environ. Toxicol. Chem. 20, 2383e2396. Di Toro, D.M., McGrath, J.A., Hansen, D.J., Berry, W.J., Paquin, P.R., Mathew, R., Wu, K.B., Santore, R.C., 2005. Predicting sediment metal toxicity using a sediment biotic ligand model: methodology and initial application. Environ. Toxicol. Chem. 24, 2410e2427. EPA, U.S., 2002. Toxicity Relationship Analysis Program (TRAP), Version 1.0. Washington, DC. Ferguson, L., Grattan, S.R., 2005. How salinity damages citrus: osmotic effects and specific ion toxicities. Horttechnology 15, 95e99. Gopalapillai, Y., Hale, B., Vigneault, B., 2013. Effect of major cations (Ca2þ, Mg2þ,   Naþ, Kþ) and anions (SO2 4 , Cl , NO3 ) on Ni accumulation and toxicity in aquatic plant (Lemna minor L.): implications for Ni risk assessment. Environ. Toxicol. Chem. 32, 810e821. He, E., Qiu, H., Van Gestel, C.A.M., 2014. Modelling uptake and toxicity of nickel in solution to Enchytraeus crypticus with biotic ligand model theory. Environ. Pollut. 188, 17e26. Hochman, Z., Dang, Y.P., Schwenke, G.D., Dalgliesh, N.P., Routley, R., McDonald, M., Daniells, I.G., Manning, W., Poulton, P.L., 2007. Simulating the effects of saline and sodic subsoils on wheat crops growing on vertosols. Aust. J. Agric. Res. 58, 802e810. ISO, 1993. Soil Quality e Determination of the Effects of Pollutants on Soil Flora e Part 1: Method for the Measurement of Inhibition of Root Growth. ISO 11269e1. Janssen, C.R., De Schamphelaere, K., Heijerick, D., Muyssen, B., Lock, K., Bossuyt, B., Vangheluwe, M., Van Sprang, P., 2000. Uncertainties in the environmental risk assessment of metals. Hum. Ecol. Risk Assess. 6, 1003e1018. Kinraide, T.B., 1999. Interactions among Ca2þ, Naþ and Kþ in salinity toxicity: quantitative resolution of multiple toxic and ameliorative effects. J. Exp. Bot. 50, 1495e1505. Kinraide, T.B., 2006. Plasma membrane surface potential jPM as a determinant of ion bioavailability: a critical analysis of new and published toxicological studies and a simplified method for the computation of plant jPM. Environ. Toxicol. Chem. 25, 3188e3198. Kopittke, P.M., Blamey, F.P.C., Kinraide, T.B., Wang, P., Reichman, S.M., Menzies, N.W., 2011. Separating multiple, short-term, deleterious effects of saline solutions on the growth of cowpea seedlings. New. Phytol. 189, 1110e1121. Kopittke, P.M., Menzies, N.W., Wang, P., McKenna, B.A., Wehr, J.B., Lombi, E., Kinraide, T.B., Blamey, F.P.C., 2014. The rhizotoxicity of metal cations is related to their strength of binding to hard ligands. Environ. Toxicol. Chem. 33, 268e277. Lexmond, T.M., 1980. The effect of soil-pH on copper toxicity to forage maize grown under field conditions. Neth. J. Agric. Sci. 28, 164e183. Li, B., Zhang, H.T., Ma, Y.B., McLaughlin, M.J., 2011. Influences of soil properties and leaching on nickel toxicity to barley root elongation. Ecotoxicol. Environ. Saf. 74, 459e466. Li, B., Zhang, X., Wang, X.D., Ma, Y.B., 2009. Refining a biotic ligand model for nickel toxicity to barley root elongation in solution culture. Ecotoxicol. Environ. Saf. 72, 1760e1766. Lin, Y., 2014. Development and Validation of a Terrestrial Biotic Ligand Model (TBLM) for Nickel and Copper Toxicities to Barley Root Elongation (Ph.D. dissertation). University of Delaware, p. 321. Lock, K., Criel, P., De Schamphelaere, K.A.C., Van Eeckhout, H., Janssen, C.R., 2007a. Influence of calcium, magnesium, sodium, potassium and pH on copper toxicity to barley (Hordeum vulgare). Ecotoxicol. Environ. Saf. 68, 299e304. Lock, K., De Schamphelaere, K.A.C., Becaus, S., Criel, P., Van Eeckhout, H., Janssen, C.R., 2006. Development and validation of an acute biotic ligand model (BLM) predicting cobalt toxicity in soil to the potworm Enchytraeus albidus. Soil Biol. Biochem. 38, 1924e1932. Lock, K., De Schamphelaere, K.A.C., Becaus, S., Criel, P., Van Eeckhout, H., Janssen, C.R., 2007b. Development and validation of a terrestrial biotic ligand model predicting the effect of cobalt on root growth of barley (Hordeum vulgare). Environ. Pollut. 147, 626e633. Lock, K., Van Eeckhout, H., De Schamphelaere, K.A.C., Criel, P., Janssen, C.R., 2007c. Development of a biotic ligand model (BLM) predicting nickel toxicity to barley (Hordeum vulgare). Chemosphere 66, 1346e1352. Munns, R., 2002. Comparative physiology of salt and water stress. Plant Cell Environ. 25, 239e250. Oorts, K., Ghesquiere, U., Swinnen, K., Smolders, E., 2006. Soil properties affecting the toxicity of CuCl2 and NiCl2 for soil microbial processes in freshly spiked soils. Environ. Toxicol. Chem. 25, 836e844. Plette, A.C.C., Nederlof, M.M., Temminghoff, E.J.M., van Riemsdijk, W.H., 1999. Bioavailability of heavy metals in terrestrial and aquatic systems: a quantitative approach. Environ. Toxicol. Chem. 18, 1882e1890. Ponizovsky, A.A., Metzler, D.M., Allen, H.E., Ackerman, A.J., 2006a. The effect of

Y. Lin et al. / Environmental Pollution 202 (2015) 41e49 moisture content on the release of organic matter and copper to soil solutions. Geoderma 135, 204e215. Ponizovsky, A.A., Thakali, S., Allen, H.E., Di Toro, D.M., Ackerman, A.J., 2006b. Effect of soil properties on copper release in soil solutions at low moisture content. Environ. Toxicol. Chem. 25, 671e682. Ponizovsky, A.A., Thakali, S., Allen, H.E., Di Toro, D.M., Ackerman, A.J., Metzler, D.M., 2008. Nickel partitioning in acid soils at low moisture content. Geoderma 145, 69e76. Rooney, C.P., Zhao, F.J., McGrath, S.P., 2007. Phytotoxicity of nickel in a range of European soils: influence of soil properties, Ni solubility and speciation. Environ. Pollut. 145, 596e605. Schecher, W.D., McAvoy, D.C., 1992. MINEQLþ e a software environment for chemical equilibrium modeling. Comput. Environ. Urban Syst. 16, 65e76. Siebielec, G., Chaney, R.L., Kukier, U., 2007. Liming to remediate Ni contaminated soils with diverse properties and a wide range of Ni concentration. Plant Soil 299, 117e130. Smith, S.R., 1994. Effect of soil-pH on availability to crops of metals in sewage sludge-treated soils .2. Cadmium uptake by crops and implications for human dietary-intake. Environ. Pollut. 86, 5e13. Sparks, D.L., 2003. Environmental Soil Chemistry. Academic Press, London, UK. Steenbergen, N.T.T.M., Iaccino, F., De Winkel, M., Reijnders, L., Peijnenburg, W.J.G.M., 2005. Development of a biotic ligand model and a regression model predicting acute copper toxicity to the earthworm Aporrectodea caliginosa. Environ. Sci. Technol. 39, 5694e5702. Tavakkoli, E., Rengasamy, P., McDonald, G.K., 2010. The response of barley to salinity stress differs between hydroponic and soil systems. Funct. Plant Biol. 37, 621e633. Thakali, S., Allen, H.E., Di Toro, D.M., Ponizovsky, A.A., Rooney, C.P., Zhao, F.J., McGrath, S.P., 2006. A terrestrial biotic ligand model. 1. Development and application to Cu and Ni toxicities to barley root elongation in soils. Environ. Sci. Technol. 40, 7085e7093. Tipping, E., 1998. Humic ion-binding model VI: an improved description of the interactions of protons and metal ions with humic substances. Aquat. Geochem.

49

4, 3e48. Tipping, E., Rieuwerts, J., Pan, G., Ashmore, M.R., Lofts, S., Hill, M.T.R., Farago, M.E., Thornton, I., 2003. The solid-solution partitioning of heavy metals (Cu, Zn, Cd, Pb) in upland soils of England and Wales. Environ. Pollut. 125, 213e225. Villagarcia, M.R., Carter, T.E., Rufty, T.W., Niewoehner, A.S., Jennette, M.W., Arrellano, C., 2001. Genotypic rankings for aluminum tolerance of soybean roots grown in hydroponics and sand culture. Crop Sci. 41, 1499e1507. Wang, P., Kinraide, T.B., Smolders, E., Zhou, D.-M., Menzies, N.W., Thakali, S., Xia, W.W., Hao, X.-Z., Peijnenburg, W.J.G.M., Kopittke, P.M., 2013. An electrostatic model predicting Cu and Ni toxicity to microbial processes in soils. Soil Biol. Biochem. 57, 720e730. Wang, P., Kinraide, T.B., Zhou, D.M., Kopittke, P.M., Peijnenburg, W.J.G.M., 2011a. Plasma membrane surface potential: dual effects upon ion uptake and toxicity. Plant Physiol. 155, 808e820. Wang, P., Kopittke, P.M., De Schamphelaere, K.A.C., Zhao, F.-J., Zhou, D.-M., Lock, K., Ma, Y.-B., Peijnenburg, W.J.G.M., McGrath, S.P., 2011b. Evaluation of an electrostatic toxicity model for predicting Ni2þtoxicity to barley root elongation in hydroponic cultures and in soils. New Phytol. 192, 414e427. Wang, P., Zhou, D.M., Peijnenburg, W.J.G.M., Li, L.Z., Weng, N.Y., 2010. Evaluating mechanisms for plant-ion (Ca2þ, Cu2þ, Cd2þ or Ni2þ) interactions and their effectiveness on rhizotoxicity. Plant Soil 334, 277e288. Weng, L.P., Lexmond, T.M., Wolthoorn, A., Temminghoff, E.J.M., Van Riemsdijk, W.H., 2003. Phytotoxicity and bioavailability of nickel: chemical speciation and bioaccumulation. Environ. Toxicol. Chem. 22, 2180e2187. Weng, L.P., Wolthoorn, A., Lexmond, T.M., Temminghoff, E.J.M., Van Riemsdijk, W.H., 2004. Understanding the effects of soil characteristics on phytotoxicity and bioavailability of nickel using speciation models. Environ. Sci. Technol. 38, 156e162. Yermiyahu, U., Nir, S., Ben Hayyim, G., Kafkafi, U., Kinraide, T.B., 1997. Root elongation in saline solution related to calcium binding to root cell plasma membranes. Plant Soil 191, 67e76. Yilmaz, D.D., 2007. Effects of salinity on growth and nickel accumulation capacity of Lemna gibba (Lemnaceae). J. Hazard. Mater. 147, 74e77.