Environ. Sci. Technol. 2008, 42, 7634–7642
Sorption of Ciprofloxacin and Oxytetracycline Zwitterions to Soils and Soil Minerals: Influence of Compound Structure ANTHONY J. CARRASQUILLO,† GREGORY L. BRULAND,‡ ALLISON A. MACKAY,§ AND D H A R N I V A S U D E V A N * ,† Department of Chemistry, Bowdoin College, Brunswick, Maine, Department of Natural Resources and Environmental Management, University of Hawai’i, Ma˜noa, Honolulu, Hawaii, and Department of Civil and Environmental Engineering, University of Connecticut, Storrs, Connecticut
Received May 20, 2008. Revised manuscript received July 22, 2008. Accepted July 24, 2008.
Oxytetracycline (OTC) zwitterions sorbed to a greater extent than ciprofloxacin (CIP) zwitterions onto goethite and soils with moderate-to-low effective cation exchange capacities (ECEC < 10 cmolc/kg) because adjacent pairs of hydroxyl groups on the OTC molecule (absent in CIP) facilitated greater surface complexation to soil metal oxides and aluminosilicate edge sites. CIP sorbed to a higher extent than OTC onto aluminosilicates and onto soils with “high” ECEC values (>10 cmolc/kg). The sorption of heterocyclic compounds structurally similar to CIP indicated that both positive charge localization on the cationic amine and the extent of charge delocalization to the heterocyclic ring influenced molecular orientation within the montmorillonite interlayers, van der Waals interactions, and the potential for sorption. The sorption of compounds structurally similar to OTC revealed that greater positive charge localization on the cationic amine facilitated sorption to montmorillonite, whereas ortho substituted anionic and cationic groups on a zwitterionic molecule resulted in unfavorable Coulombic interactions between the anionic moiety and the negatively charged surface and hindered sorption. Thus, greater CIP zwitterion sorption to aluminosilicates and “high” ECEC soils resulted from greater distance between the anionic and cationic groups, which maximized Coulombic attraction to the surface.
Introduction Study of the fate and transport of veterinary antibiotics has been motivated by several factors, including the high volume of veterinary antibiotic sales in the United States and Europe (1), the large extent (up to 72%) of excretion of unmetabolized antibiotics (2), the detections of veterinary antibiotics at subinhibitory concentrations in soils, surface waters, and ground waters in the U.S. and Europe (3, 4), and the global increase in antibiotic resistant strains of microbes (5). The * Corresponding author phone: (207)725-3548; fax: (207)725-3017; e-mail:
[email protected]. † Bowdoin College. ‡ University of Hawai’i. § University of Connecticut. 7634
9
ENVIRONMENTAL SCIENCE & TECHNOLOGY / VOL. 42, NO. 20, 2008
potential for antibiotic sorption and desorption within soil systems plays a key role in their environmental fate. Current understanding of fluoroquinolone and tetracycline sorption to soils and pure phase minerals is relatively well advanced, and similarities in the sorption behavior of these compounds are well understood. For example, both compound classes exist as cations, zwitterions, and anions at environmentally relevant pH values (pH 4-8) (Figure S1, Supporting Information) and display pH dependent sorption to soils (6-8). In addition, the extent of fluoroquinolone and tetracycline sorption to soils was found to be strongly influenced by the soils’ cation exchange capacity (CEC) (6, 8, 9) and, to some extent, by the soils’ metal oxide content (8, 9). Studies of sorption onto model soil components (such as metal oxides and aluminosilicate clays) have revealed similarities in the mechanisms of tetracycline and fluoroquinolone sorption; both compound classes were found to sorb onto aluminosilicate clays via cation exchange and cation bridging (electrostatic attraction or complexation of the anionic moiety on the antibiotic zwitterion to exchangeable cations) (10-13) and to participate in surface complexation or ligand exchange with metal oxides via the bonding of carboxyl or hydroxyl groups to surficial metal ions (7, 14-16). Comparison of two independent studies of oxytetracycline (a high-use tetracycline) and ciprofloxacin (a high-use fluoroquinolone) zwitterion sorption onto the same soils pointed to a large number of soils exhibiting greater extent of oxytetracycline sorption and a smaller group of soils exhibiting greater ciprofloxacin sorption (8, 9). This cursory analysis led to our hypothesis that differences in the extents of fluoroquinolone and tetracycline sorption to the same soil matrix result from key differences in compound structure and availability of surface sites for sorption via a particular mechanism. Our goal was to determine structural criteria responsible for differences in oxytetracycline and ciprofloxacin zwitterion sorption to soils and soil minerals. Objectives included understanding the influence of cationic amine group structure on the extent of cation exchange to aluminosilicates and the effect of ligand group (e.g., carboxyl, hydroxyl) identity and positioning on the extent of complexation to surfacebound metal ions.
Experimental Section To achieve our objectives, earlier independent measurements of ciprofloxacin and oxytetracycline zwitterion sorption to the same set of 30 soils (8, 9) were reanalyzed to identify the differences in the soil factors influencing the extents of ciprofloxacin and oxytetracycline sorption and to hypothesize preferred sorption mechanisms. Subsequently, specific structural factors, such as electronic, delocalization, proximity, solvation, and steric effects, were evaluated by determining the extent of ciprofloxacin and oxytetracycline sorption to pure phase minerals (montmorillonite, kaolinite, and goethite) and sorption of structurally similar compounds (related in structure to ciprofloxacin and oxytetracycline) to montmorillonite using methods detailed below. Materials. Detailed information on the sorbents (kaolin, Ca-montmorillonite, and goethite), sorbates (ciprofloxacin, oxytetracycline, and structurally similar compounds), and rationale for compound selection are found in the Supporting Information. Chemical structures and abbreviations for all compounds are found in Figure S1. All solutions were prepared in deionized water (18 MΩ), and all glassware was acid-washed prior to use. 10.1021/es801277y CCC: $40.75
2008 American Chemical Society
Published on Web 09/05/2008
Sorption-Desorption Studies. Sorption. Experimental conditions were designed to achieve between 15-95% sorption, reach supernatant concentrations above the limits of quantitation, and, where relevant, achieve an extent of sorption well below the sorbent CEC. Sorption was measured in triplicate using predetermined solid loadings (1 g/L, 5 g/L, or 10 g/L) and an initial solute concentration of 0.1 mM (for CIP and OTC sorption onto kaolinite and montmorillonite) or 5.0 × 10-5 M (all other compounds and for CIP and OTC sorption to goethite) in either a 10 mM chloroacetate (pH 3) or PIPES (pH 7) solution. CIP and OTC sorption to montmorillonite, kaolinite, and goethite was measured at pH ∼ 7, at which zwitterion concentrations were dominant in aqueous solution (Figure S1). The sorption of structurally similar compounds possessing a pKa of 4.5 or greater (which accounts for all compounds except 1PP) (Table 1) were measured at pH 3, at which the protonated (cationic) amine was dominant. Sorption of 1PP was measured at pH ∼7 (between the two pKas of 4.49 and 8.63 (17)), at which the mono cation was the dominant species. An appropriate mass of solid was transferred into preweighed 15 mL polypropylene centrifuge tubes, and then 10 mL of appropriately buffered solute solution was added. The same solution was also added to solid-free centrifuge tubes that served as blanks. Based on previous studies of reaction kinetics of our test compounds (11, 13, 17-19), all reactors were set to rotate end over end for 24 h in the dark (to prevent photodegradation). Reactors were then centrifuged (3500 rpm for 45 min at 15 °C), the supernatant was decanted and filtered, and solute concentrations were determined by high performance liquid chromatograph and diode array detection (HPLC methods for all compounds are found in the Supporting Information, Table S1). The concentration sorbed, Cs (mol/kg), was calculated by subtracting solute concentrations in the solid-containing rectors, Cw (mol/L), from solute concentrations in the corresponding solid-free reactors and then normalizing for the mass of solid in the reactor. The distribution coefficient or equilibrium sorption constant, Kd (L/kg), defined as Cs/Cw was then calculated. For OTC and CIP alone, an equilibrium sorption constant normalized to aqueous zwitterion concentrations, K′d, was derived as outlined in the Supporting Information. In all cases, average Kd and K′d values and related uncertainties (one standard deviation) were reported based on triplicate sorption experiments. Comparison of the extent of sorption observed in our studies with previous studies of the same compounds utilizing similar experimental conditions (7, 11, 12, 14, 15, 19, 20) indicated that our Cs values were within the linear range of the respective sorption isotherms, providing justification for our use of single-point Kd values. Desorption. Desorption experiments were performed for the aluminosilicate systems to evaluate the reversibility of the sorption process. Tetrabutylammonium hydroxide (TAH) solution at pH 10 was used as the desorbing agent (8, 10) because TAH possesses a permanent charge, whereas the solutes examined are either neutral or anionic at pH 10. As such, TAH facilitates desorption by replacing the solute molecules on the aluminosilicate surface. Following the sorption experiments, the supernatant was completely decanted, and the mass of solute in the fluid entrained between solid particles was calculated. Next, 5 mL of 10 mM TAH solution at pH 10 was added to the reactor, and the wet-solid was resuspended and set to rotate end over end for 24 h in the dark. Reactors were then centrifuged, and the supernatant was decanted, filtered, and analyzed using HPLC-DAD. The percent of the solute desorbed from the surface was then determined. The average desorption for structurally similar compounds was found to be 62 ( 20% and pointed to the reversibility of the sorption process; compounds exhibiting lower extents of sorption showed
limited reversibility with our desorption techniques. Desorption of CIP and OTC from pure phase minerals under similar experimental conditions has been documented previously (7, 12, 20) and, hence, was not repeated in this study. Quantum Chemical Methods. Charge and area of individual atoms within a molecule, molecular volume, and molecular dimensions were determined using Spartan ’06, Version 1.12 (Wave function Inc.). Energy minimizations were performed to ensure that all calculations were carried out with the molecule in its most energetically stable configuration. Subsequently, the Austin Model 1 or AM1 semiempirical method was employed (ground state, equilibrium geometry, subject to symmetry) for all computations. The area per atom was calculated based on a space-filling model. Two calculated parameters relevant to this study are cationic amine charge/area and heterocyclic ring charge/ area. The cationic amine moiety was defined as the amine nitrogen atom and all hydrogens attached to that nitrogen. The heterocyclic ring moiety (used for the purpose of evaluating charge delocalization into the heterocyclic ring) was defined as the five carbon atoms (or four carbons in the case of 1PP) comprising the ring; the heterocyclic cationic nitrogen was excluded as it was included in the amine moiety. The charge and area of a specific moiety were calculated by summing the charges and areas of individual atoms that comprised the moiety. Subsequently, charge/area was derived by dividing the total charge of the moiety by the total area of the moiety. Electrostatic potential maps were created based on AM1 calculations to obtain a visual representation of the relative charge densities over the entire molecule. Statistical Analyses. Results from two previous studies that individually examined CIP and OTC sorption to the same set of 30 soils were reanalyzed; OTC sorption was measured at pH 5.5 (9) and CIP sorption at five pH values between pH 3 and pH 8 (8). Principal component analysis (PCA) was conducted in these earlier studies with a data set comprised of 29 soils (Burton soil omitted to ensure normality of the data set), 14 or 16 soil properties, and Kd values for CIP or OTC sorption. Here, additional PCA (PC-ORD, MJM Software Design) were designed to probe the differences in soil factors influencing OTC and CIP sorption. PCA was conducted on the entire set of 29 soils and a reduced set of 22 soils. The seven soils omitted in the “reduced” data set were found to have an undue influence on the principal component ordination space (as explained in the Results and Discussion section). Of the data available from the previous work, Kd values at pH 5.5 and pH 7 were selected for OTC and CIP sorption, respectively, because the zwitterion concentrations were predominant (OTC: 98% and CIP: 85%) and anion concentrations were minimal (OTC and CIP: 1%) at these pH values (Figure S1). K′d values were then calculated for the comparison of CIP and OTC zwitterion sorption. Next, twelve soil properties, including descriptors of texture and surface area (%clay, %sand, surface area), exchange capacity (effective cation exchange capacity (ECEC), exchangeable (Ex) Ca, Ex Mg, Ex Na, Ex K, and Ex Al), organic carbon content (% OC), and iron and aluminum oxide content (sum of dithionite-citrate-bicarbonate extractable aluminum and iron (DCBAl+Fe) and acid ammonium oxalate extractable aluminum and iron (AAOAl+Fe)), were selected from the earlier characterization data (9). The significance of the principal component axes was tested with the broken-stick eigenvalue test (21). Values of the above-mentioned properties for all 30 soils are found in the Supporting Information. A seminal study by Tolls (22) found that data treatment with organic carbon normalization was conceptually inappropriate for veterinary pharmaceuticals (such as CIP and OTC) because a number of hydrophobicity independent mechanisms such as cation exchange/bridging, surface VOL. 42, NO. 20, 2008 / ENVIRONMENTAL SCIENCE & TECHNOLOGY
9
7635
7636
9
ENVIRONMENTAL SCIENCE & TECHNOLOGY / VOL. 42, NO. 20, 2008
+1 +1 +/-0
+1
7500 ( 580
64 ( 2.5 2800 ( 149 19480 ( 1100
+1
6700 ( 830
+/-0 +/-0
+1
380 ( 1.0 2 ( 0.12 769 ( 90
+1 +1 +1 +1 +1
Charge
22 ( 0.3 18 ( 1.3 47 ( 0.83 75 ( 2.2 150 ( 2.1
Kd or K′db
5.09 8.63e 8.76f
6.36
6.3
4.54 9.44d
5.63
4.61 5.04 4.7 5.26 5.1
amine pKac
3.90 5.00
6.67 0.06 0.47
0.26
0.34 -0.81 -0.85
-0.03
-0.02
16.76 17.21 17.18
11.12
10.81
0.09 0.20 0.19
-0.03
-0.10
Ciprofloxacin and Related Structures
-1.47 -4.18
0.52
0.26 0.23 0.22
0.29
0.30
0.35 0.24
0.26
0.27 0.23 0.25
0.28 0.28 0.27 0.27
Oxytetracycline and Related Structures -0.63 29.33 -0.06 0.28 -0.52 29.33 -0.04 0.28 -0.05 16.54 0.06 0.27 -0.01 16.54 0.08 0.27 0.37 7.09 0.22 0.26
N
H2i
cationic amineh area (Å) H1i
logDow,c,g
elemental charge
h
0.28 0.28
H3i
+ 0.62 + 0.66 + 0.65
+ 0.25
+ 0.20
+ 0.41 + 0.71
+ 0.52
+ 0.78 + 0.79 + 0.60 + 0.61 + 0.48
amine group total chargeh
+ 3.42 × 10-3 + 3.85 × 10-3 - 1.66 × 10-2 - 2.37 × 10-2 -2.42 × 10-2
+ 2.28 × 10-2 + 3.71 × 10-2 + 3.84 × 10-2 + 3.81 × 10-2
ringh
+ 1.81 × 10-2
+ 1.06 × 10-1 + 1.41 × 10-1
+ 7.76 × 10-2
+2.66 × 10-2 + 2.69 × 10-2 + 3.61 × 10-2 + 3.70 × 10-2 + 6.77 × 10-2
cationic amine
charge/areah
a Compound structures found in Figure S1. b K′d listed for CIP and OTC, Kd for all other compounds. c Estimated from SciFinder Scholar unless otherwise noted. d Tavares et al., 1994 (31). e Zhuang and Huang, 2007 (17). f Vasquez et al., 2000 32. g Dow values estimated at pH of the sorption experiment. h Spartan calculations described in the Methods section. i H1, H2, and H3 shown in Figure S1.
5,6,7,8-tetrahydroquinoline (5678THQ) 5,6,7,8-tetrahydroisoquinoline (5678THIQ) 1,2,3,4-tetrahydroquinoline (1234THQ) 1-phenylpiperazine (1PP) ciprofloxacin (CIP)
aniline p-toluidine N-methylaniline (MA) N-methyl-p-toluidine (MT) N,N-dimethylaniline (DMA) N,N-dimethyl-p-toluidine (DMT) 2-(dimethylamino)-5methylbenzenesulfonate (DMT-S) oxytetracycline (OTC)
compounda
TABLE 1. Kd Values for Sorption to Montmorillonite and Properties of All Test Compounds
FIGURE 1. Sorption of CIP and OTC onto (a) soils with ECEC < 10 cmolc/kg and K′dOTC between 12,500 and 2400 L/kg, (b) soils with ECEC < 10 cmolc/kg and K′dOTC < 2400 L/kg, and (c) soils with ECEC > 10 cmolc/kg. complexation, and hydrogen bonding were involved in sorption to soils. As such, organic carbon normalized sorption coefficients (Koc/Kom) were not evaluated in our statistical analyses.
Results and Discussion Comparison of CIP and OTC Sorption to Soils. Comparison of the extent of CIP and OTC zwitterion sorption (K′d) onto the same soils provided insights into the influence of soil properties and compound structure not directly evident from previous studies of individual compounds (8, 9). The extent of OTC zwitterion sorption was greater than CIP zwitterion sorption (Figure 1a,b) for most soils examined. However, a small group of 8 soils exhibited a higher extent of CIP sorption (Figure 1c). Notably, these 8 soils possessed significantly higher (two tailed t test, unequal variances; p < 0.05) ECEC values (mean: 31.6 cmolc/kg; standard error (SE): 5) and an excess of cation exchange/bridging sites (mmolc g mmolsolute) than the remaining 22 soils (mean ECEC: 3.8 cmolc/kg; SE: 0.5). Assuming ECEC to be a measure for both cation exchange and cation bridging sites, it appears that the CIP molecular structure allowed for a greater extent of zwitterion sorption only when cation exchange/bridging sites were not limited. In all other cases, the OTC structure appeared to be better suited for zwitterion sorption to soils. Examination of PCA biplots depicting the position and loading of soil property and CIP and OTC K′d vectors in ordination space provided additional insight into the distinct role of soil properties on the nature and extent of CIP and OTC sorption. Analysis of the complete data set (29 soils) revealed similar loading of soil property vectors (lines on the biplot) and samples (individual soils represented as points on the biplot) in ordination space for the PCA of soil properties and K′d,CIPpH7 (Figure 2a) and for the PCA of soil properties and K′d,OTCpH5.5
(Figure 2b) (detailed PCA description in the Supporting Information). In addition, both PCAs showed a grouping of 7 soils in the high ECEC/high surface area zone of the biplot (circled in Figure 2a,b) and a clustering of the remaining 22 soils in the low ECEC/low surface area zone of the biplot. These seven soils all possessed an ECEC > 20 cmolc/kg and were comprised primarily of soils of the Vertisol order, with the exception of one soil (an Alfisol). To examine how this grouping influenced the positioning of soil descriptor vectors and soils in ordination space, we repeated the PCA with only 22 soils, excluding the 7 soils noted above. The PCA of these 22 soils (reduced data set) resulted in a wider distribution of soils within ordination space and showed a slightly modified loading of the soil descriptors (Figure 2c, Figure S4). Descriptors of soil texture (%clay and SA) loaded strongly on axis 2 rather than axis 1. There was, however, little difference with respect to the loading of descriptors of soil cation exchange capacity (which remained on axis 1) and iron aluminum oxide content (on axis 2) (Figure 2c, Figure S4). Differences in the loading of K′d,CIPpH7 and K′d,OTCpH5.5 in the ordination space of PCAs with 29 and 22 soils revealed the influence of ECEC on CIP zwitterion sorption and the soil metal oxide content on OTC zwitterion sorption. K′d,CIP pH7 exhibited a stronger loading on axis 1 than axis 2 in the PCA with 29 (loading of 0.34 on axis1 and of 0.25 on axis 2) and 22 soils (loading of 0.39 on axis 1 and 0.12 on axis 2), emphasizing as previously noted (8) the stronger correlation between ECEC and K′d,CIPpH7 and the weaker correlation between soil metal oxide content and K′d,CIPpH7. Together, these results indicate that ECEC is an important determinant of the magnitude of K′d,CIPpH7 regardless of availability of cation exchange/bridging sites (i.e., the magnitude of ECEC). Interestingly, the position of K′d,OTCpH5.5 in ordination space was distinctly different in the PCA with 29 vs 22 soils. K′d,OTCpH5.5 loaded between the axis 1 and 2 in the PCA of 29 VOL. 42, NO. 20, 2008 / ENVIRONMENTAL SCIENCE & TECHNOLOGY
9
7637
FIGURE 2. Principal component analysis (PCA) of (a) soil properties and K′d,CIPpH7 with 29 soils, (b) soil properties and K′d,OTCpH5.5 with 29 soils, and (c) soil properties and K′d,OTCpH5.5 with 22 soils. Note: PCA of soil properties and K′d,CIPpH7 with 22 soils is shown in Figure S4 (Supporting Information). soils with double the loading on the iron and aluminum oxide content axis (axis 2 loading: 0.46) as compared to the texture/ECEC axis (axis 1 loading: 0.2) (Figure 2b) and pointed to the influence of both ECEC and soil metal oxide content on the magnitude of K′d,OTCpH5.5 (as observed previously (9)). The distinct difference in the PCA of 22 soils was that K′d,OTCpH5.5 loaded primarily on the texture/aluminum and iron oxide content axis (0.42 loading on axis 2) and exhibited a weak loading on the exchange capacity axis (0.04 on axis 1) (Figure 2c). Therefore, it appeared that the magnitude of K′d,OTCpH5.5 was primarily influenced by soil metal oxide content and texture in soils with moderate-to-low ECEC values (
