Advances in freshwater risk assessment: improved accuracy of dissolved organic matter-metal speciation prediction and rapid biological validation

Highlights

• Some commonly used bioavailability models for risk assessment are inaccurate.

• Introduces a rapid, accurate approach to optimize speciation-based risk models.

• Demonstrates rapid model validation using a whole-cell bioreporter.

• Describes first regional demonstration project for streamlined ERA in freshwater.

Abstract

Speciation modeling of bioavailability has increasingly been used for environmental risk assessment (ERA). Heavy metal pollution is the most prevalent environmental pollution issue globally, and metal bioavailability is strongly affected by its chemical speciation. Dissolved organic matter (DOM) in freshwater will bind heavy metals thereby reducing bioavailability. While speciation modeling has been shown to be quite effective and is validated for use in ERA, there is an increasing body of literature reporting problems with the accuracy of metal-DOM binding in speciation models. In this study, we address this issue for a regional-scale field area (Lake Tai, with 2,400 km² surface area and a watershed of 36,000 km²) where speciation models in common use are not highly accurate, and we tested alternative approaches to predict metal-DOM speciation/bioavailability for lead (Pb) in this first trial work. We tested five site-specific approaches to quantify Pb-DOM binding that involve varying assumptions about conditional stability constants, binding capacities, and different components in DOM, and we compare these to what we call a one-size-fits-all approach that is commonly in use. We compare model results to results for bioavailable Pb measured using a whole-cell bioreporter, which has been validated against speciation models and is extremely rapid compared to many biological methods. The results show that all of the site-specific approaches we use provide more accurate estimates of bioavailability than the default model tested, however, the variation of the conditional stability constant on a site-specific basis is the most important consideration. By quantitative metrics, up to an order of magnitude improvement in model accuracy results from modeling active DOM as a single organic ligand type with site-specific variations in Pb-DOM conditional stability constants. Because the biological method is rapid and parameters for site-specific tailoring of the model may be obtained via high-throughput analysis, the approach that we report here in this first regional-scale freshwater demonstration shows excellent potential for practical use in streamlined ERA.

Introduction

According to reports from the Pure Earth Institute and partners and other scientific studies (Clemens and Ma, 2016; Pure Earth/Green Cross, 2015; RoyChowdhury et al., 2018; Tchounwou et al., 2012), heavy metal pollution is the world’s biggest pollution problem. Concomitant to the intensification of heavy metal pollution globally, the long-used approaches to environmental risk assessment (ERA) are comparatively unwieldy (e.g. US Environmental Protection Agency, USEPA, 2001), and there is an increasing urgency for accurate and streamlined ERA (Janssen et al., 2003; Yang et al., 2018; Zhang et al., 2017). Heavy metals are introduced to freshwater environments by many different industrial activities, namely mining, refining ores, smelting, legacy sources such as leaded gasoline, fertilizer industries, tanneries, battery manufacture, paper industries, pesticides and wastewater (Pastircakova, 2004; RoyChowdhury et al., 2018). In terms of heavy metal toxicants, lead (Pb) is ranked as the most abundant heavy metal pollutant in the world today (Pure Earth/Green Cross 2015, also, historically, see Tong et al., 2000). In 2016, the Institute for Health Metrics and Evaluation (IHME) estimated that Pb exposure accounted for 540,000 deaths worldwide due to its long-term effects on health (WHO, 2019) and that the global annual costs of childhood Pb exposure from cognitive defects is as much as 81 trillion dollars (Attina and Trasande, 2013; Grandjean and Bellange, 2017), most of which is borne by low- and middle-income countries. Therefore, new solutions for ERA concerning the old and persistent problem of Pb pollution are imperative.

One often-utilized approach in ERA involves collecting large amounts of environmental monitoring data and making risk projections based on total concentrations, which do not consistently agree well with toxic response (USEPA, 2003a). Biological effects, such as toxic response, are a consequence of pollutant bioavailability in the environment, hence bioavailability has been an increasing focus of ERA (USEPA, 2003b), particularly for heavy metals. Metal bioavailability in turn is determined by chemical speciation, which dictates that metals may be present in chemically different forms, i.e. as free ion or complexes with inorganic (e.g., Cl^-, SO₄^2-, CO₃^2-) and natural organic ligands; speciation is thus governed by water chemistry (Clifford and McGeer, 2010; Sisombath, 2014). One of the most successful approaches to bioavailability-based ERA is the Biotic Ligand Model (BLM), which was developed to predict metal toxicity to aquatic organisms using simple and inexpensively obtained water chemistry parameters as inputs (Di Toro et al., 2001; Sander et al., 2015; USEPA, 2007a, USEPA, 2007b). Many studies have demonstrated that, subsequent to biological validation, the BLM is a time-saving approach to obtain optimized information for streamlined ERA (see, for example, Lock et al., 2007; Nys et al., 2014; Rüdel et al., 2015; Wang and Song, 2019). The BLM is calculated in speciation or toxicity mode, the latter being used to predict toxicity for any given water chemical composition, however, the fundamental model (or precursor to toxicity) is based on chemical speciation of a metal. Results from the BLM reflect how metal complexation with inorganic and organic ligands will decrease metal toxicity by decreasing the concentration of toxic metal free ions, which is generally accepted to be the relevant quantity that leads to toxic response in bioavailability-based ERA (Nys et al., 2014; USEPA, 2007a, USEPA, 2007b, 2016).

Speciation-based models used in ERA need to be validated. Validation is conducted by comparing model results to results from biological measurements. Many biological techniques have been used for validation including, as examples, reduction of free metal via binding of copper (Cu) and silver (Ag) to Oncorhynchus mykiss gills (Smith et al. 2017), acute lethality tests of Daphnia pulex in response to cadmium (Cd) (Clifford and McGeer, 2010), and suppression of elongation of Hordeum vulgare root in response to trivalent chromium (Cr) (Song et al., 2014). Model validation using such approaches can be very accurate, however the approaches are time consuming and relatively labor intensive in terms of cultivating and maintaining test organisms (Heijerick et al., 2005; Nys et al., 2014; Smith et al., 2017). Typically, validation relies on toxicity (indirect), however, many investigators now begin to use more advanced and rapid techniques (such as the whole-cell bioreporter method used herein) that can measure speciation as well. Bioreporters are organisms that are genetically engineered to generate a signal and give a “report” on target substances and are capable of producing dose-dependent signals in response to target analytes (Belkin, 2003; Kessler et al., 2012; van der Meer and Belkin, 2010). In recent years, bioreporter techniques have attracted increasing attention for their role in measuring the bioavailability of pollutants in the environment (Al-Anizi et al., 2014; van der Meer and Belkin, 2010; Wells, 2012). There are now many reports utilizing bioreporters that are specifically responsive to heavy metals to assess the bioavailability and toxicity of Cu, Pb, arsenic (As), cobalt (Co), nickel (Ni) and zinc (Zn) (Jia et al., 2016; Magrisso et al., 2009; Ndu et al., 2012; Trang et al., 2005; Yoon et al., 2016a, Yoon et al., 2016b), including reports demonstrating agreement between bioreporter and speciation model/BLM results (An, et al., 2012.; Ndu et al., 2012; Thakali et al., 2006; Zhang et al., 2017).

With respect to the crucial need to develop new streamlined techniques for ERA of Pb, we recently validated the use of a Pb-sensitive bioreporter against a speciation model commonly used in ERA and for BLM modeling (Zhang et al., 2017). Controlled lab studies (e.g. as used in Zhang et al., 2017) have shown that the bioreporter technique can be used as a reliable tool for validation work when the prediction of metal bioavailability by the speciation model is accurate; this accuracy may be readily achieved for such studies because they utilize well-characterized solution components. In contrast, in natural aquatic systems, metal speciation, hence metal bioavailability, may be highly variable because it is primarily determined by complexation with DOM, which reduces bioavailability and hence risk, (Baken et al., 2011; Yamashita and Jaffe', 2008; Zhang et al., 2017, 2020), however, speciation models do not always adequately predict metal-DOM binding (Ahmed et al., 2014). Accuracy may suffer due to the variable nature of DOM (e.g. see Ahmed et al., 2014; Ndungu, 2012). As the composition of DOM is variable, so too is the metal binding strength and capacity of DOM (Zhang et al., 2020). The Stockholm Humic Model (SHM) (Gustafsson, 2001) and the Windemere Humic Acid Model (WHAM) (Tipping, 1994 and 1998) are two common modeling approaches to quantify metal-DOM binding and attendant metal bioavailability with readily available software. To perform speciation calculations, these models crucially rely on conditional stability constants (K_cond), which originate from thermodynamic quantities and are a quantity described as a constant that reflects the strength of the interaction of metal-DOM. Such models assume that HA and fulvic acid (FA) are the main binding components of DOM and that a single K_cond describes metal binding with phenolic and carboxylic-acid type sites within HA and FA, respectively (Gustafsson, 2001; Tipping, 1994; Tipping et al., 1998). Due to the use of an invariant K_cond, we have referred to this as a “one-size-fits-all” approach (Zhang et al., 2020). Since the composition and binding strength of DOM in natural water can vary substantially in some cases (e.g. see Mostofa et al., 2013; Zhang et al., 2014; Zhang et al., 2020), though the one-size-fits-all approach has been very successful, it will not always be amenable to accurate calculation of speciation and assessment of risk.

Our primary aim in the work reported here was to find a simple and rapid way to alter extant speciation models to more accurately represent metal-DOM binding in cases where the one-size-fits-all models fail to provide accurate agreement with biological measurements. It is important to have an approach that is fit-for-purpose regarding ERA in freshwater, and we chose a large lacustrine system for this work since lakes simultaneously represent an endpoint reflecting processes across whole catchments and play an important role with respect to the effect of biodiversity on ecosystem stability (Lévêque, 2001). Lake Tai (hereafter designated as Taihu, after 太湖 in Chinese) was selected as the target study area as it is large enough (>2,400 km²) to reflect regional-scale processes, has a complex aquatic ecosystem and has a long history of anthropogenic impacts (Qin et al., 2007; Sun and Mao, 2008). For this work, we wanted to study a field setting that would enable regional-scale study in a setting that we know from experience provides enough variability in DOM (K_cond and binding capacity, C_L) to test the effects of variability on model outputs (Zhang et al., 2020). Taihu is located in the southeast part of the Yangtze River Delta, which is a highly developed area in China. As China’s third largest freshwater lake, Taihu is important for a variety of purposes including as a drinking water source, for flood control, for tourism and for aquaculture (Gong and Lin, 2009; Qin et al., 2007). While the particular field site we chose is well-suited to this work, the general approach would be extensible to most freshwater environments. The objectives of this study are 1) to develop and test different approaches to introducing site-specific variation in speciation models of Pb-DOM binding, 2) to evaluate the performance of the different approaches to speciation modeling using the previously validated bioreporter technique and 3) to examine which factors are most important in controlling the accuracy of site-specific models. We find that the optimized speciation model results are much more accurate in their agreement with bioreporter results, a result that serves as a good demonstration for how ERA might be at once customized/site-specific and streamlined in freshwater settings.