Clarithromycin and Tetracycline Binding to Soil Humic Acid in the

Department of Soil Science, University of Wisconsin, Madison, Wisconsin 53706, United States. Environ. Sci. Technol. , 2016, 50 (18), pp 9933–9942. ...
0 downloads 9 Views 2MB Size
Article pubs.acs.org/est

Clarithromycin and Tetracycline Binding to Soil Humic Acid in the Absence and Presence of Calcium Iso Christl,† Mercedes Ruiz,‡ J.R. Schmidt,§ and Joel A. Pedersen*,‡,§,⊥ †

Institute of Biogeochemistry and Pollutant Dynamics, ETH Zurich, 8092 Zürich, Switzerland Environmental Chemistry and Technology Program, University of Wisconsin, Madison, Wisconsin 53706, United States § Department of Chemistry, University of Wisconsin, Madison, Wisconsin 53706, United States ⊥ Department of Soil Science, University of Wisconsin, Madison, Wisconsin 53706, United States ‡

S Supporting Information *

ABSTRACT: Numerous ionizable organic micropollutants contain positively charged moieties at pH values typical of environmental systems. Describing organic cation and zwitterion interaction with dissolved natural organic matter requires explicit consideration of the pH-dependent speciation of both sorbate and sorbent. We studied the pH-, ionic strength-, and concentration-dependent binding of relatively large, organic cations and zwitterions (viz., the antibiotics clarithromycin and tetracycline) to dissolved humic acid in the absence and presence of Ca2+ and evaluated the ability of the NICA-Donnan model to describe the data. Clarithromycin interaction with dissolved humic acid was well described by the model including the competitive effect of Ca 2+ on clarithromycin binding over a wide range of solution conditions by considering only the binding of the cationic species to low proton-affinity sites in humic acid. Tetracycline possesses multiple ionizable moieties and forms complexes with Ca2+. An excellent fit to experimental data was achieved by considering tetracycline cation interaction with both low and high protonaffinity sites of humic acid and zwitterion interaction with high proton-affinity sites. In contrast to clarithromycin, tetracycline binding to humic acid increased in the presence of Ca2+, especially under alkaline conditions. Model calculations indicate that this increase is due to electrostatic interaction of positively charged tetracycline-Ca complexes with humic acid rather than due to the formation of ternary complexes, except at very low TC concentrations.



INTRODUCTION A variety of important organic and organometallic contaminants contain moieties that are positively charged over at least part of the pH range typical of natural waters and soil solutions. These moieties include aliphatic and alicyclic amines, aromatic amines, aromatic heterocyclic nitrogens, and organometallic compounds. In addition to possessing basic functionalities, many of these compounds contain acidic moieties leading to complex, pH-dependent speciation, and for some compounds, the formation of zwitterions (e.g., fluoroquinolone and tetracycline antibiotics, glyphosate, hydroxyisoquinones) and net anions carrying a cationic functional group (e.g., tetracycline antibiotics). Humic substances influence organic contaminant phase distribution, transport, and bioavailability.1−3 Depending on their structure, organic contaminants may engage in electrondonor/electron−acceptor and electrostatic interactions with humic substances, as well as van der Waals interactions.3−5 Some contaminants may also covalently bind with humic substances.6−8 Electrostatic interactions are expected to contribute substantially to the binding of ionizable organic © XXXX American Chemical Society

contaminants to humic substances. Humic substances are net polyanionic over the entire pH range typical of natural environments. Their net negative charge increases with pH due to deprotonation of primarily carboxyl and phenolic functional groups. Organic cation and zwitterion sorption to humic substances exhibits a strong pH-dependence, reflecting the speciation of both the sorbent and sorbate.9−17 Accurate description of organic cation and zwitterion interaction with dissolved humic substances therefore requires explicit consideration of the pHdependent charge of both. Single-parameter linear free energy relationships based on log Kow fail to predict ionizable organic compound sorption to natural organic matter (NOM).18 The size and type of charged group (primary vs quaternary amine) was shown to be predictive for the sorption affinity of alkylbenzylamine and alkylbenzylammonium compounds to Received: February 5, 2016 Revised: July 6, 2016 Accepted: July 20, 2016

A

DOI: 10.1021/acs.est.5b04693 Environ. Sci. Technol. XXXX, XXX, XXX−XXX

Article

Environmental Science & Technology

Figure 1. Speciation as a function of pH, skeletal formulas and molecular electrostatic potentials (MEPs) of (a) clarithromycin and (b) tetracycline, and (c) MEP for sulfathiazole cation. Speciation was calculated for an ionic strength of 0.01 M. Macroscopic dissociation constants (pKa) for clarithromycin and tetracycline were taken from Nakagawa et al.33 and Mitscher,39 respectively. Molecular electrostatic potentials were calculated along the ρ = 0.03 e/Å3 electron density isosurface and are superimposed over ball-and-stick structures. Atoms in the ball-and-stick structures are color-coded as follows: white, H; gray, C; blue, N; red, O; and yellow, S. Abbreviations: CLA0, neutral clarithromycin; CLA-H+, clarithromycin cation; STA+, sulfathiazole cation; TC-H3+, tetracycline cation; TC-H2±, tetracycline zwitterion; TC-H+/2‑, tetracycline monoanion; TC2−, tetracycline dianion.

soil organic matter.3 Simple sorption models that rely on binding constants to describe organic cation or zwitterion binding may not accurately describe sorption to NOM over wide pH and concentration ranges.19 Previous attempts to describe the pH-dependence of organic cation and zwitterion interaction with dissolved humic acid using a discrete log K spectrum model predict general trends,9,20 but in some cases do not adequately describe sorption as a function of pH and are subject to arbitrariness in parameter assumptions with respect to humic acid protonation.14,17 Furthermore, inorganic cations are ubiquitous in natural waters and can compete with organic cations for binding to NOM. Compared to previous attempts,14 more complex models that explicitly account for electrostatic effects are required and need to be tested for their ability to satisfactorily predict the influence of inorganic cations differing in valency. Several models have been developed to describe inorganic cation sorption to humic substances.21 Among inorganic cation binding models, the nonideal competitive adsorption (NICA)Donnan model has been shown to successfully describe cation binding over large pH and metal cation concentration ranges and to accurately depict cation competition for binding sites in humic substances. In the NICA-Donnan model, the binding of cations to specific sites in humic substances is described using a continuous, bimodal affinity distribution. Nonspecific electrostatic accumulation in the humic acid is accounted for using a

Donnan approach, in which humic acids are considered a gellike phase having a uniformly distributed potential. The negative charge of the Donnan phase arising from deprotonation of humic acids is compensated by counterions. Recently, the NICA-Donnan model was shown to describe sulfathiazole binding to Leonardite humic acid over wide pH and sorbate concentration ranges.19 This represented the first application of the NICA-Donnan model to an organic compound. However, competition between organic and inorganic cations was not addressed. Cationic sulfathiazole interaction with negatively charged humic acid moieties dominated sulfathiazole binding even at pH values at which the cation was a minor solution species. The apparently successful application of the NICA-Donnan model to sulfathiazole interaction with humic acid suggests that the ability of the model to describe organic cation/zwitterion sorption to NOM warrants further investigation. Few subsequent studies have attempted to apply the NICA-Donnan model to describe organic cation binding to NOM.22,23 At present, information is limited on the strength of organic cation binding to humic acids, the degree to which binding affinity can be related to the type of the cationic moiety, and the extent to which other structural moieties (e.g., those allowing H-donor/ acceptor or π−π interactions) contribute to cation affinity for humic acids. Furthermore, for organic cations with multiple ionizable functional groups, the relative influence of cationic B

DOI: 10.1021/acs.est.5b04693 Environ. Sci. Technol. XXXX, XXX, XXX−XXX

Article

Environmental Science & Technology

from deprotonation of various functional groups in humic acids. The negative charge of the humic phase is compensated by countercation accumulation within the Donnan volume, VD (L·kg−1), according to the empirical relationship

and anionic moieties on binding is unclear. The pH-dependent speciation of some organic contaminants is substantially more complex than that of the inorganic cations for which the NICADonnan model was originally developed. Here, we examine the applicability of the NICA-Donnan model to describe the pH- and concentration-dependent sorption of relatively large, organic cations and zwitterions to dissolved humic acid in the absence and presence of Ca2+. We investigate whether the NICA-Donnan model can successfully describe the interaction of (1) clarithromycin (CLA) and (2) tetracycline (TC) with Elliott soil humic acid (ESHA) chosen as a representative terrestrial humic acid. Clarithromycin bears a single ionizable moiety and exists as a cation at pH < pKa (Figure 1). As CLA does not form aqueous calcium complexes, Ca2+ is expected to compete with CLA binding to ESHA. Tetracycline contains three ionizable functionalities and is present as a cation at pH < pKa1, a zwitterion at pKa1 < pH < pKa2, a net monoanion exhibiting one positive and two negative charges at pKa2 < pH < pKa3, and a dianion at pH > pKa3 (Figure 1). Calcium cations can form positively charged complexes with TC and are therefore expected to both promote and compete with TC binding to ESHA depending on the solution conditions. We first used new and previously published data for CLA and TC binding to ESHA14,17 to derive binding constants for both compounds. Using the calibrated model, effects of Ca2+ on CLA and TC binding to ESHA were then predicted and compared with experimental data. In this paper, we use the term binding to refer to cation/zwitterion interaction with specific anionic moieties in humic substances without implying covalent bond formation. Nonspecific electrostatic interaction refers to accumulation in the humic phase to compensate the negative electrostatic potential.

log VD = b(1 − log I ) − 1

where I is ionic strength and b is an empirical parameter describing the change in the Donnan volume as a function of ionic strength. In the NICA-Donnan model, the amount of cation i specifically bound to humic acid, Qi, is given by [∑i (K1,̃ iCi)n1,i ]p1 (K1,̃ iCi)n1,i · Qi = ·Q · n1,H max1,H ∑i (K1,̃ iCi)n1,i 1 + [∑i (K1,̃ iCi)n1,i ]p1 n1, i

+

n2, i n2,H

·Q max2,H·

[∑i (K̃ 2, iCi)n2,i ]p2 (K̃ 2, iCi)n2,i · ∑i (K̃ 2, iCi)n2,i 1 + [∑i (K̃ 2, iCi)n2,i ]p2

where Ci denotes the concentration in the Donnan phase, the subscripts j = 1 and 2 correspond to low and high protonaffinity sites (primarily carboxylic and phenolic sites, respectively), nj,H are heterogeneity parameters for proton binding, nj,i describe cation-specific heterogeneity parameters, Qmaxj,H represent the maximum proton site densities of the humic substance for each distribution (mol·kg−1), the K̃ j,i are median affinity distribution constants, and the pj represent the widths of the distributions and describe the intrinsic heterogeneity of the humic substance. The total amount of cation i associated with humic acid is calculated as the sum of cations Qi bound to specific sites and the amount of cation i accumulated in the Donnan phase due to nonspecific electrostatic accumulation. Based on the pH-dependence of CLA binding to ESHA reported previously,17 only positively charged CLA-H+ was assumed to interact with ESHA. For TC binding, the triply protonated TC-H3+ cation and the doubly protonated TC-H2± zwitterion, that is, the dominant species at acidic to circumneutral pH were considered to interact with ESHA because of the decrease in TC binding at alkaline conditions. For fitting NICA-Donnan model parameters specific to antibiotic binding (viz. the affinity constants K̃ j,i and ionspecific heterogeneity parameters nj,i) log-transformed data on CLA and TC binding were used. Model parameters were optimized using the programs ECOSAT and FIT to account for the aqueous speciation of all species during fitting.26,27 The Akaike Information Criterion corrected for small sample size (AICC) was used to select among competing assumptions on interactions considered in modeling.28 Based on the best fits, the effect of Ca2+ present in solution was predicted. General NICA-Donnan model parameters for ESHA and ion-specific parameters for H+ and Ca2+ binding to ESHA were taken from Christl.29 In addition to NICA-Donnan model equilibria (vide supra), all calculations considered the complete pH- and ionic strength-dependent aqueous speciation of all components present in solution (viz. background electrolytes, CLA, and TC including Ca−TC complex formation).30 Molecular Modeling. Density functional theory calculations on clarithromycin, tetracycline, and sulfathiazole monomers were conducted using Gaussian 0931 in conjunction with the B3LYP density functional. Monomer geometries were optimized in the gas phase using the 6-31G(d) basis; the molecular electrostatic potential (MEP) was calculated from the density at the B3LYP/cc-pVDZ level in the presence of a polarizable continuum solvation model (PCM).32 All reported



MATERIALS AND METHODS Experimental Data. Experimental data on CLA interaction with dissolved ESHA (IHSS, No. 1S102H) were taken from Sibley and Pedersen.17 This published data set comprises concentration- (pH 6.5 and ionic strength, I = 0.01 M), pH- (I = 0.01 M), and ionic strength-dependent (pH 6.5, I = 0.02 and 0.22 M) data on CLA binding to ESHA at a total ESHA concentration, [ESHA]tot ∼ 0.07 g·L−1. New experimental data on CLA binding to ESHA were collected to provide a more robust data set for modeling. New data presented here include CLA binding isotherms for pH 8.7 (±0.1) and 9.0 (±0.1) measured at I = 0.01 M. Furthermore, TC binding isotherms for pH 3.2 (±0.1), 4.3 (±0.0), 6.4 (±0.1), and 9.1 (±0.1) were recorded at I = 0.01 M, and pH-dependent binding to ESHA was investigated at I = 0.01 and 0.1 M and [TC]total = 36−96 nM. The new experimental data were collected at [ESHA]tot ∼ 0.07 g·L−1. To further extend the data sets to trace concentrations and to investigate the effect of Ca2+ on binding of antibiotics at trace levels, 3H-labeled CLA and TC were used to determine CLA and TC binding to ESHA at [Ca2+]tot = 0− 10 μM. Due to a significant variability in published data on CLA solubility in aqueous media (see Supporting Information, SI), we additionally determined CLA solubility at pH 12 by equilibrating an excess of solid CLA for 48 h in a 0.01 M phosphate solution. Experimental details on the acquisition of new data are given in the SI. NICA-Donnan Modeling. A detailed description of the NICA-Donnan model has been presented elsewhere.24,25 Briefly, the model depicts humic acids as a gel-like phase exhibiting a uniformly distributed negative potential resulting C

DOI: 10.1021/acs.est.5b04693 Environ. Sci. Technol. XXXX, XXX, XXX−XXX

Article

Environmental Science & Technology MEP values are maxima along the ρ = 0.03 e/Å3 electron density isosurface, corresponding approximately to the molecular van der Waals radius. Specific binding interactions were examined by modeling interaction of cationic monomers with a benzoic acid anion, yielding an overall neutral dimer. Dimer binding energies were calculated with respect to isolated ions (including PCM solvation) at the B3LYP/cc-pVTZ/PCM level, including augment functions on O atoms.32 Reported energies were counterpoise corrected using the corresponding gas phase dimer at the same geometry.



RESULTS AND DISCUSSION Clarithromycin. Clarithromycin consists of a 14-membered lactone ring with two attached deoxy sugars (Figure 1a). The ionizable moiety is the dimethylamine group on one deoxy sugar (pKa = 8.9).33 The clarithromycin cation CLA-H+ is thus the dominant species over the pH range typical of natural waters and soil solutions. Binding of clarithromycin to ESHA is characterized by an envelope-shaped pH-dependence and pronounced ionic-strength dependence (Figure 2a). Isotherm data display linearity over a large concentration range on a log− log plot and exhibit a slope close to unity (Figure 2b). The pH-, ionic strength-, and concentration-dependence of clarithromycin binding to ESHA were well described (R2 = 0.9422) by considering only CLA-H+ binding to low protonaffinity Q1 sites (e.g., carboxylates), that is, by fitting only two parameters, the affinity constant K̃ 1,CLA‑H+ and the ion-specific heterogeneity parameter n1,CLA‑H+ (Figure 2 and Table 1). Assuming that CLA-H+ binds to only high proton-affinity Q2 sites resulted in a poorer fit to the data (R2 = 0.6151). When both Q1 and Q2 sites were included, the model attributed binding almost exclusively to Q1 and the fit was not improved relative to the assumption of binding to Q1 only (Table 1). Our modeling result indicates that humic functional groups with a low affinity for protons control the binding of clarithromycin in two ways. First, binding to Q1 sites is calculated to dominate clarithromycin interaction with ESHA (SI Figure S1). Second, nonspecific electrostatic accumulation of positively charged CLA-H+ in the Donnan phase due to the potential from mainly deprotonated Q1 sites becomes quantitatively important when pH decreases to acidic conditions (SI Figure S1). The negligible contribution of CLA0 to overall sorption is consistent with the high polarity of clarithromycin and the absence of aromatic groups (i.e., π−π interactions with humic acid not possible). Furthermore, binding of clarithromycin to humic acid under alkaline conditions is controlled by the deprotonation of its amine group. Deprotonation decreases the fraction of CLA present as the cationic species (Figure 1a) and limits the total amount of dissolved CLA due to the low solubility of CLA0 in aqueous medium, which we found to be 0.083 ± 0.006 mg·L−1 (log Cwsat = −7.0, Cwsat in molar units). This value is within the range of reported solubilities (see section S2 in SI).33−36 The fitted median affinity constant for CLA-H+ binding to Q1 sites, log K̃ 1,CLA‑H+, was 0.3 (Table 2). This value is substantially lower than the median proton-affinity constant for these sites (log K̃ 1,H = 2.2). The binding affinity of CLA-H+ is similar to values reported for metal cations having relatively low affinities for Q1 sites.37 Therefore, the presence of metal cations is expected to strongly affect clarithromycin binding to humic acid because they can readily compete for binding sites. The fitted value of nCLA‑H+ (1.04 ± 0.02) is close to unity. Koopal et al.38 showed that the ion-specific parameters ni relate directly to the

Figure 2. Binding of clarithromycin (CLA) to Elliott soil humic acid (ESHA) as a function of (a) pH and ionic strength and (b) as a function of nonsorbed dissolved CLA at pH 6.5 (±0.1), 8.7 (±0.1), and 9.0 (±0.1). QCLA denotes the amount of total CLA bound to ESHA, CCLA is the amount of total nonsorbed CLA in solution at equilibrium. Experimental data on pH-dependent binding in (a) and concentration-dependent data at pH 6.5 in (b) were taken from Sibley and Pedersen17 and correspond to total ESHA concentrations of 65−74 mg·L−1. Isotherms at pH 8.7 and 9.0 were recorded at 72 (±1) mg·L−1 ESHA. Fits are based on NICA-Donnan model calculations considering binding of cationic CLA-H+ to low proton-affinity sites of ESHA (Q1 sites) only. For the fit to the entire data set, R2 = 0.9422.

slope of the cation binding isotherm at low cation loadings. Comparison with NICA-Donnan affinity constants recently reported for sulfathiazole (STA) binding to Leonardite humic acid19 indicates cationic sulfathiazole exhibits considerably higher affinity for Q1 sites of humic acid (log K̃ 1,STA+ = 3.25) than does cationic clarithromycin; K̃ 1,STA+ is 3 orders of magnitude larger than K̃ 1,CLA‑H+. Tetracycline. Tetracycline contains four fused rings (designated A through D, Figure 1b) and multiple ionizable moieties, leading to more complex speciation than is the case for clarithromycin. Three macroscopic dissociation constants have been reported relevant to the pH range for most natural environments.39 The most acidic pKa value (3.30) is assigned to the tricarbonylamide system on ring A. The second dissociation constant (7.68) is attributed to the phenolic diketone system spanning rings B to D. The third pKa value (9.69) corresponds to the dimethylamine functionality on ring A. The pHdependence of tetracycline binding to ESHA is characterized by a maximum located in the strongly acidic pH range (i.e., at a D

DOI: 10.1021/acs.est.5b04693 Environ. Sci. Technol. XXXX, XXX, XXX−XXX

Article

Environmental Science & Technology

therefore fitted the complete data set including pH-, concentration- and ionic strength-dependent data with the NICA-Donnan model assuming that only TC-H3+ interacts with ESHA. For clarithromycin (vide supra) and sulfathiazole,19 binding to humic acid was well-described by considering interaction of only the cationic species of the organic sorbate. For tetracycline, however, this simple approach did not produce an adequate fit, even when binding to both Q1 and Q2 sites was considered (R2 = 0.6763, Table 1). The fit was substantially improved, when both TC-H3+ and the zwitterionic species, TCH2±, were assumed to interact with ESHA (R2 ≥ 0.9483, Table 1). To evaluate which pairs of interactions between TC-H3+ and TC-H2± and the binding sites Q1 and Q2 are required to obtain the best fit, we employed the Aikake information criterion (AIC), which accounts for both the number of fitted parameters and the residual sum of squares.28 Based on AIC values, the best fit was obtained when TC-H3+ was assumed to interact with Q1 and Q2 sites and TC-H2± with Q2 sites (R2 = 0.9696, Table 1). These fitting results indicate that tetracycline interacts in a more complex way with humic acids than do clarithromycin and sulfathiazole. Both cationic and zwitterionic TC interact importantly with dissolved humic acid in contrast with sulfathiazole.19 For sulfathiazole sorption, the zwitterionic species, STA±, was not relevant, even at pH values at which STA± abundance in solution surpasses that of STA+ (e.g., at pH 7, [STA±]/[STA+] > 300). The negligible contribution of zwitterionic STA± to STA binding was therefore due to its low affinity for anionic sites in humic acid. The dominant species between pH 2.1 and 7.2 is STA0. STA± is a minor species, in tautomeric equilibrium with STA0, and does not dominate solution speciation at any pH because tautomeric equilibrium limits [STA±]/[STA0] to ∼0.004. For tetracycline, the zwitterionic species is not limited by tautomeric equilibrium with a neutral species. The abundance of TC-H2± and its affinity for humic acid allows it to contribute substantially to tetracycline binding to ESHA at neutral and alkaline pH (SI Figure S2). For the best fit, median affinities for TC-H3+ and TC-H2± binding to ESHA were log K̃ 1,TC‑H3+ = 2.7, log K̃ 2,TC‑H3+ = 7.6,

Table 1. Quality of NICA-Donnan Model Fits for Describing Clarithromycin (CLA) and Tetracycline (TC) Binding to Elliott Soil Humic Acid in the Absence of Ca2+ Assuming Different Binding Interactions fits to full data sets

number of fitted parameters

R2

Δia

2 2 4

0.9422 0.6151 0.9422

0 174 4

2 4 4 6

0.6763 0.9380 0.9483 0.9483

384 114 84 88

6

0.9696

0

6

0.9686

5

8

0.9696

4

Clarithromycin (n = 92) a) CLA-H+ binds to Q1 a) CLA-H+ binds to Q2 b) CLA-H+ binds to Q1 and Q2 Tetracycline (n = 166) a) TC-H3+ binds to Q1 b) TC-H3+ binds to Q1 and Q2 c) TC-H3+ and TC-H2± bind to Q1 d) TC-H3+ binds to Q1 and Q2 and TC-H2± binds to Q1 e) TC-H3+ binds to Q1 and Q2 and TC-H2± binds to Q2 f) TC-H3+ binds to Q1 and TC-H2± binds to Q1 and Q2 g) TC-H3+ and TC-H2± bind to Q1 and Q2

AICC difference, Δi = AICCi − minAICC, where AICCi is the Akaike Information Criterion (AIC) for model i corrected for small sample size and minAICC is the smallest AIC value over the set of models compared, and AICC = AIC + 2k(k + 1)/(n − k − 1); AIC = 2k + n[ln (RSS/n)]; k = number of parameters in the statistical model; n = number of observations; RSS = residual sum of squares. The plausibility that model i represents the best approximating model of the candidate set decreases as Δi increases: models with Δi < 2 have substantial support, those with 4 < Δi < 7 have substantially less support, and those with Δi > 10 have negligible support.28 a

pH value close to the lowest pKa of tetracycline; Figure 3a). Ionic strength-dependent data exhibit an intersection in the slightly acidic pH range. Isotherms at pH 3.2, 4.3, 6.4, and 9.1 plotted on a log−log scale indicate a linear relationship between dissolved and ESHA-associated tetracycline with slopes close to unity (Figure 3b). The pH-dependence of tetracycline binding to ESHA indicated a pivotal role of cationic TC-H3+. As a first step, we

Table 2. NICA-Donnan Model Parameters for Describing Clarithromycin (CLA) and Tetracycline (TC) Binding to Elliott Soil Humic Acid (ESHA)a general parameters for ESHA −0.12 low affinity site, Q1 3.0 0.73 ion-specific parameters

Donnan parameter b site densities Qj,H (mol·kg−1) heterogeneity parameters pj

protons H+ calcium Ca2+ clarithromycin (R2 = 0.9422) CLA-H+b tetracycline (R2 = 0.9696) TC-H3+ TC-H2±b

high affinity site, Q2 3.8 0.23

log K̃ 1

n1

log K̃ 2

n2

2.2

0.89

8.2

0.88

−1.7

0.50

0.3 ± 0.1

1.04 ± 0.02

2.7 ± 0.1

0.94 ± 0.02

7.6 ± 0.1 6.2 ± 0.1

0.80 ± 0.06 0.88 ± 0.03

a

General parameters and ion-specific parameters for H+ and Ca2+ binding to ESHA were taken from Christl.29 Parameter values in bold were determined from fit to the experimental CLA and TC data. bAffinities are corrected by the respective pKa1 of CLA and TC to directly give the affinities of the species CLA-H+ and TC-H2±. E

DOI: 10.1021/acs.est.5b04693 Environ. Sci. Technol. XXXX, XXX, XXX−XXX

Article

Environmental Science & Technology

(K̃ 1,TC‑H3+ = 2.7) was indispensable for a good fit at low pH. This result indicates that the negative charge present on the zwitterion strongly influences binding to ESHA. Our molecular modeling results show the negatively charged moieties on tetracycline decrease the maximum MEP in the vicinity of the cationic dimethylamino group from 6.9 V for the cation to 3.7 V for the zwitterion (Figure 1b). This reduces electrostatic attraction to the net negatively charged humic acid; the lower affinity constants reflect the resulting decrease in binding. In the net monoanionic species, TC-H1+/2−, the dimethylamino group remains positively charged and an additional moiety is negatively charged. For TC-H1+/2−, a further decrease in affinity for Q1 and Q2 sites can be expected relative to TC-H2±. We cannot exclude that TC-H1+/2− also binds to humic acid at very alkaline pH based on our modeling results. Nonetheless, the quantitative contribution of TC-H1+/2− is expected to be very small. Binding of Antibiotic Cations to Low Proton-Affinity Sites. We compared the molecular electrostatic potentials and energies of binding to the benzoate carboxylate group for CLAH+ and TC-H3+ with those for the sulfathiazole cation, STA+, to help rationalize the observed differences in median affinity constants for low proton-affinity Q1 sites. Affinities of the antibiotic cations for Q1 sites in humic acids (ESHA for CLAH+ and TC-H3+, Leonardite humic acid for STA+)19 increased in the order CLA-H+ (log K̃ 1,CLA‑H+ = 0.03 ± 0.01) ≪ TC-H3+ (log K̃ 1,TC‑H3+ = 2.7 ± 0.1) < STA+ (log K̃ 1,STA+ = 3.3 ± 0.1). The cationic moiety in sulfathiazole is an aromatic primary amine (anilinium cation), while those in clarithromycin and tetracycline are aliphatic tertiary amines. Molecular modeling suggests that part of the difference in affinities is due to the type of amine. The methyl groups of the tertiary amines serve to hinder the approach of the nitrogen atom to a humic carboxylate moiety. As a result, the carboxylate group encounters a substantially less positive MEP for the clarithromycin and tetracycline tertiary amines than it does for the less sterically hindered primary amine in sulfathiazole. The MEPs calculated at the van der Waals radius are significantly less positive for clarithromycin (7.5 V) and tetracycline (6.9 V) than for sulfathiazole (9.5 V) (Figure 1), resulting in weaker electrostatic interactions with anionic moieties. Furthermore, explicit calculations of sulfathiazole, clarithromycin, or tetracycline binding to a benzoate anion (as a simple model for Q1 sites in humic acid) show the formation of a stable hydrogen bond between the benzoate and the amine.40 As a result of both increased electrostatic interaction, and possibly also the decreased steric hindrance about the primary (vs tertiary) amine, the sulfathiazole complex is bound more strongly (125 kJ·mol−1) than that for clarithromycin (48 kJ·mol−1) or tetracycline (53 kJ·mol−1). The magnitudes of the dimer binding energies are undoubtedly too large because the PCM underestimates solvation effects; however, the trend in binding and the explanation are valid. The magnitudes of the MEPs and estimated energies of binding with the benzoate anion lead to the expectation that the affinities of CLA-H+ and TC-H3+ for Q1 sites should be similar if the only interaction operative was binding of the protonated tertiary amine with deprotonated carboxylate sites. The large difference in the K̃ 1 values for CLA-H+ and TC-H3+ (more than 2 orders of magnitude) indicates that additional types of interactions must be considered to explain the divergence in their affinities for Q1 sites. Determination of

Figure 3. Binding of tetracycline (TC) to Elliott soil humic acid (ESHA) at ionic strengths (I) of 0.01 and 0.1 M as a function of (a) pH and (b) nonsorbed dissolved TC concentration at pH 3.2 (±0.1), 4.3 (±0.0), 6.4 (±0.1), and 9.1 (±0.1). QTC denotes the amount of total TC bound to ESHA, CTC is the amount of total nonsorbed TC in solution at equilibrium. The data were obtained at total ESHA concentrations of 63−74 mg·L−1. NICA-Donnan model fits are based on binding of cationic TC-H3+ to low and high proton-affinity sites of ESHA (Q1 and Q2 sites) and binding of zwitterionic TC-H2± to Q2 sites. For the fit to the entire data set, R2 = 0.9696.

and log K̃ 2,TC‑H2± = 6.2 (Table 2). The fitted values of the ionspecific heterogeneity parameters nj ranged from 0.80 to 0.94 (Table 2) and were close to the respective heterogeneity parameters fitted for proton binding.29 Corresponding to the best fit, the speciation of tetracycline bound to ESHA reveals the following pH-dependent binding sequence: TC-H3+−Q1 dominates at pH < 6.5, TC-H3+−Q2 at pH 6.5−8.2, and TCH2±−Q2 at pH > 8.2 (SI Figure S2). This ranking reflects the abundance of TC-ESHA species at I = 0.01 M. Nonspecific tetracycline accumulation in the Donnan phase was found to be quantitatively not relevant. Even for the pH 3.2 isotherm data, only 0.2% of total tetracycline associated with ESHA was attributed to nonspecific accumulation in the Donnan volume. In both TC-H3+ and TC-H2±, the dimethylamino group attached to ring A is positively charged. Our fitting results indicate that the affinity of TC-H3+ for Q2 is more than one logunit higher than that of TC-H2± (Table 2). When both TC-H3+ and TC-H2± were assumed to bind to Q1 and Q2 sites, parameter optimization tended to decrease TC-H2± affinity for Q1 to levels at which the contribution of this species to total binding was minor (≤8%), whereas the species TC-H3+−Q1 F

DOI: 10.1021/acs.est.5b04693 Environ. Sci. Technol. XXXX, XXX, XXX−XXX

Article

Environmental Science & Technology the precise mechanisms of CLA-H+ and TC-H3+ interaction with Q1 sites requires further study. Nevertheless, some comments can be made on factors impacting the magnitude of the affinity constants. Both compounds can engage in hydrogen bonding via hydroxyl groups. Tetracycline is more rigid and polarizable than CLA. Unlike CLA, TC has the ability to engage in π−π interactions and possesses multiple nonester carbonyl groups that can serve as H-bond acceptors in Hbonding with NOM.14,41 Such interactions may reinforce TC binding to Q1 sites, leading to a larger affinity constant than for CLA. The bulkiness of CLA may hinder its approach to anionic Q1 sites, reducing the strength of interaction. Effect of Ca2+ on CLA and TC Binding to ESHA. In natural systems, the presence of inorganic cations (e.g., Ca2+, Mg2+) is expected to affect organic cation binding to humic substances because the former may be present at much higher concentrations. We tested the effect of Ca2+ on the binding of CLA and TC to ESHA at two pH values (5.7 and 9.1) These pH values were chosen because (1) in the slightly acidic pH range Ca2+ is bound to a large extent to Q1 sites but does not influence the speciation of the antibiotics to a major extent and (2) in the alkaline pH range, TC forms a positively charged complex with Ca2+ (vide infra) while clarithromycin does not. The pH conditions selected allowed us to observe pronounced differences experimentally and to challenge the model. Our experimental data show that total calcium concentrations of 10−6 M and higher affect the binding of micromolar amounts of clarithromycin to ESHA at pH 5.6 and 9.1 (Figure 4a). At both pH values, an increase in total calcium diminished clarithromycin binding to ESHA demonstrating that Ca2+ competes with clarithromycin for binding to humic acid sites even at low Ca2+ activities. A recent study on the pH- and ionic strength-dependent binding of Ca2+ to ESHA29 revealed that nonspecific binding contributed most to Ca2+ binding at I = 0.01 M, when pH was in the acidic range and Ca2+ activities exceeded 10−6. Specific binding of Ca2+ was favored by increasing pH and attributed exclusively to interactions with low proton-affinity sites (Q1 sites) in ESHA.29 Consequently, clarithromycin and Ca2+ compete mainly for binding to sites with low proton affinities within humic acid. Competitive effects due to nonspecific accumulation may be minor unless Ca2+ is present as a major cation significantly contributing to or controlling ionic strength. Consonant with our experimental data, the NICA-Donnan model predicted an increase in total calcium concentration at pH 9.1 to reduce clarithromycin binding to ESHA (Figure 4a). At pH 5.6, however, the model underestimated the effect of Ca2+ at [Ca2+]total = 10−5 M, because Ca2+ binding is increasingly attributed to nonspecific binding in the model (13% at [Ca2+]total = 10−6 M and 51% at [Ca2+]total = 10−5 M), which does not result in a distinct decrease in clarithromycin binding. In contrast to clarithromycin, increasing total calcium concentrations promoted tetracycline binding to ESHA (Figure 4b). This effect was more pronounced at pH 9.0 than pH 5.7. Tetracycline can form several complexes with Ca2+,30 which become increasingly important as pH increases (Figure 5a). Among the Ca−TC complexes, TC-Ca22+ is net positively charged and may associate with negatively charged humic acids due to nonspecific electrostatic accumulation. The calibrated NICA-Donnan model that did not consider ternary TC−Ca− ESHA complexes predicted elevated Ca2+ concentrations to slightly promote TC binding (Figure 4b). The calculations allowing TC-Ca22+ to accumulate in the Donnan volume

Figure 4. Effect of Ca on binding of (a) clarithromycin (CLA) and (b) tetracycline (TC) to Elliott soil humic acid (ESHA) at ionic strength, I = 0.01 M as a function of total calcium concentration and pH. QCLA and QTC represent the amount of total CLA and TC bound to ESHA, respectively. Symbols represent experimental data recorded at total concentrations of 67−69 mg·L−1 ESHA and 1.1 to 1.9 × 10−6 M CLA or 3.4 to 6.2 × 10−7 M TC, respectively. Solid lines were calculated based on NICA-Donnan model fits considering binding of cationic CLA-H+ to low proton-affinity sites of ESHA (Q1 sites) in plot (a) and binding of cationic TC-H3+ to low and high proton-affinity sites of ESHA (Q1 and Q2 sites) and binding of zwitterionic TC-H2± to Q2 sites in plot (b). The dashed line in plot (b) corresponds to a model calculation assuming additional ternary complexes (viz. binding of TCCa22+ to ESHA; K̃ 1,TC-Ca2+ = 3.0).

suggested that nonspecific electrostatic accumulation of this complex is too weak to describe the effects of Ca2+ on tetracycline binding observed at submicromolar tetracycline concentrations. The absence of Ca2+-induced competition with tetracycline binding to ESHA can be rationalized by the following two findings. First, Q2 sites of ESHA bind TC-H3+ and TC-H2± (this study), but not Ca2+.29 Second, Q1 sites of ESHA exhibit a higher affinity for TC-H3+ than for Ca2+ (Table 2). These considerations imply that the observed enhancement of tetracycline binding to ESHA in the presence of Ca2+ may be due to the binding of TC-Ca22+ with specific sites (i.e., formation of ternary TC−Ca−ESHA complexes). We investigated with simulation the strength of such ternary complexes that would be required to cause the observed enhancement of TC binding in the presence of Ca2+. To examine this, the TCCa22+ complex was tested for binding to Q1 sites in ESHA. The total amount of tetracycline bound to ESHA did not increase considerably at pH 9.0 before TC-Ca22+ affinity for Q1 exceeded G

DOI: 10.1021/acs.est.5b04693 Environ. Sci. Technol. XXXX, XXX, XXX−XXX

Article

Environmental Science & Technology

higher concentrations, but tend to deviate at trace TC concentrations (