Environ. Sci. Technol. 2006, 40, 7236-7241
Water Hardness as a Photochemical Parameter: Tetracycline Photolysis as a Function of Calcium Concentration, Magnesium Concentration, and pH† JEFFREY J. WERNER,‡ W I L L I A M A . A R N O L D , * ,‡,§ A N D K R I S T O P H E R M C N E I L L * ,‡,| Water Resources Science Program, University of Minnesota, 1985 Buford Avenue, St. Paul, Minnesota 55108, Department of Civil Engineering, University of Minnesota, 500 Pillsbury Drive SE, Minneapolis, Minnesota 55455-0116, and Department of Chemistry, University of Minnesota, 207 Pleasant St. SE, Minneapolis, Minnesota 55455
The environmental photochemical kinetics of the antibiotic compound tetracycline were investigated. The aqueous speciation of tetracycline over a range of natural pH and water hardness values is dominated by association with Ca2+ and Mg2+ ions. The association constants necessary to calculate tetracycline aqueous speciation given knowledge of pH, [Ca2+], and [Mg2+] were measured by spectrophotometric titrations and matrix deconvolution of a series of UV-vis absorption spectra into individual component species. A series of photolysis experiments was performed under simulated sunlight, and quantum yields for the solar photolysis of each environmentally relevant species were calculated. The results indicate that the pseudofirst-order rate constant for tetracycline photolysis at varied Mg2+ and Ca2+ concentrations relevant to natural conditions can vary by up to an order of magnitude. A selfsensitization effect was observed and was accounted for by varying the initial tetracycline concentration under each set of photolysis conditions.
Introduction Pharmaceuticals and personal care products (PPCPs), compounds making up a recently recognized diverse class of pollutants, have been observed to have a widespread presence in anthropogenically impacted surface waters (1, 2). The complete ecological and public health implications of the current suite of PPCPs present in natural waters are not well understood. Because PPCPs are designed to have a physiological effect and to be recalcitrant to metabolic degradation, they are considered to pose a potential environmental threat (3, 4). Antibiotic compounds are a particularly alarming subset of the PPCP pollutant class because, in addition to potential ecotoxicity, the addition of antibiotics to environmental systems presents the possibility for the development of or †
This article is part of the Emerging Contaminants Special Issue. * Corresponding author phone: (612)625-8582; e-mail:
[email protected] (W.A.A.) or phone (612)625-0781; e-mail:
[email protected] (K.M.). ‡ Water Resources Science Program. § Department of Civil Engineering. | Department of Chemistry. 7236
9
ENVIRONMENTAL SCIENCE & TECHNOLOGY / VOL. 40, NO. 23, 2006
selection for antibiotic-resistant genes. Enhanced antibiotic resistance by these mechanisms may occur in natural water bodies where wastewater treatment plants (WWTPs) discharge effluent and also on farm fields where antibioticcontaining manure is spread. Tetracycline (TTC) antibiotics, used extensively in both human and animal medicine, have been detected in WWTP effluent and in surface waters downstream of WWTP effluent and agricultural fields (2, 5, 6). Photodegradation is expected to be a significant loss process for TTC in some surface waters. A discussion of the potential importance of photolysis to the environmental fate of pharmaceutical compounds is available in the review by Boreen et al. (7). Antibiotic compounds of the TTC class are known to be particularly unstable upon exposure to light. Several studies on the photolysis of TTC have been published (8-13), but none have been performed using both environmentally relevant conditions (aqueous solution and the solar spectrum) and a quantitative kinetic analysis, making it currently difficult to accurately predict the environmental photochemical kinetics of TTC. The products of photolysis, however, have been the subject of much investigation. The primary products of TTC photolysis are predominantly lumitetracycline and anhydrotetracycline, the ratio of which depends on the presence of oxygen, pH, and binding to metal ions (11). Anhydrotetracycline is toxic to bacteria by a different mechanism than that of TTC (14). Other products observed to result from TTC photolysis are smaller fragments no longer exhibiting the tetracycline structure (10). It is therefore expected that loss of tetracycline by photolysis will lead to a reduction of the risk toward proliferation of tetracycline resistance. TTC exhibits a complicated aqueous speciation which must be considered in the assessment of its environmental fate. The variables affecting speciation will also affect aqueous photochemistry, for each species will have its own quantum yield and light absorption properties. This concept has been demonstrated for some members of the sulfa drug class of antibiotics (15). In Figure 1, the TTC cation is shown with the acidic protons labeled according to their pKa values. It is important to note that the assignment of specific protons to pKa values in Figure 1 is for illustrative purposes only; the observed macroscopic pKa values of tetracycline are the result of microscopic equilibria for which the order of acidity for protons b and c is reversed for alternate conformations of the molecule (16, 17). By potentiometric titration, Regna et al. (18) observed the pKa values of oxytetracycline to shift significantly downward upon addition of calcium chloride, demonstrating the ability of tetracyclines to bind strongly to divalent metals. It is therefore hypothesized in the current study that the dominant dissolved TTC forms in the environment will be bound to Ca2+ and Mg2+, because these are the divalent metal cations present at the highest concentrations, and the association constants seem to be large. Support for this hypothesis comes from Beliakova et al. (13), who noted that the rate of photolysis of TTC increased when 5 mM Mg2+ was added to a buffered photolysis solution. Association with Ca2+ and Mg2+ has also been noted to affect the photolysis rates of other organic compounds. Landymore and Antia (19) attributed rapid photolysis of two pteridine compounds in seawater to enhanced deprotonation brought about by binding to Ca2+ and Mg2+. Because of the enhanced deprotonation of acidbase functional groups upon association with metal cations, it is important to note that a fourth acidic proton with a pKa of 11.8 has been observed for TTC at I ) 0.1 (20). Although 10.1021/es060337m CCC: $33.50
2006 American Chemical Society Published on Web 06/23/2006
FIGURE 1. Top: the tetracycline cation. The pKa values of each acidic proton (at I ) 0) are (a) 3.45, (b) 8.00, (c) 9.82 (this work), and (d) 12.37 (ref 20, adjusted via the Davies equation from 11.8 at I ) 0.1). Bottom: possible Ca2+/TTC and Mg2+/TTC equilibria which were considered, assuming total [TTC] is small enough to avoid multiple-ligand complexes; M ) Ca2+ or Mg2+; L ) TTC; the metalbinding equilibrium constants used in the text are defined as the stepwise association constants represented by each set of equilibrium arrows shown here. this pKa value is not environmentally relevant, it is possible that metal binding could cause it to become significantly lower, requiring the fourth deprotonation to be considered in the total speciation scheme. The equilibrium constants for the binding of the aqueous TTC conjugate bases to Ca2+ and Mg2+ ions are needed in order to assess the species-dependent processes affecting the fate of total TTC in the environment. The potentiometric determination of the Ca2+/TTC and Mg2+/TTC binding constants is frustrated by the multiple-ligand complexes which form at the necessarily high experimental TTC concentrations (21, 22). Potentiometric titration alone therefore cannot yield an answer. Martin (23), using spectrophotometric observations upon addition of metal ions, reported 1:1 metal:ligand binding constants for the Ca2+/TTC and Mg2+/TTC systems, considering only the first three pKa values in the speciation scheme. It was noted, however, that the 1:1 binding constants would be a misinterpretation of the data if a second metal-binding event occurred under the conditions of the experiment. Fluorometric observations by Schmitt and Schneider (24) provide evidence for a second metalbinding site, resulting in 2:1 Ca2+:TTC and Mg2+:TTC complexes under environmentally relevant conditions (pH 7.0-8.5), possibly involving loss of the fourth acidic proton. Considering these cited observations, the total possible speciation scheme for either the Ca2+/TTC or the Mg2+/TTC system at low total [TTC] is that shown in Figure 1. The initial goals of this study were to measure the physical constants required to calculate the aqueous speciation of TTC in the presence of environmentally relevant Ca2+ and Mg2+ concentrations. For each relevant dissolved TTC species, the quantum yield for direct photolysis under natural sunlight was then determined as well as the UV-vis decadic molar absorptivity spectrum. The values measured in the current study allow for kinetic predictions of the photolysis of TTC under natural sunlight, given knowledge of pH, [Ca2+], and [Mg2+].
Experimental Section Materials. Chemicals, suppliers, and the procedures for the preparation of stock solutions are listed in the Supporting Information. Speciation Studies. The pKa values of TTC and metalligand association constants for the Ca2+/TTC and Mg2+/ TTC systems were determined from UV-vis absorption spectra obtained during pH titrations at varying [Ca2+] and
FIGURE 2. Molar absorptivity of TTC at 400 nm as a function of pH, with no metals present (triangles) and with [Ca2+] ) 10-1.5 M (squares). The lines shown are the fit of the speciation model to the data. [Mg2+]. A description of the experimental conditions and data fitting procedure is available in the Supporting Information. Photochemical Kinetics. The direct photolysis of TTC was monitored in experiments for which the values of pH, [Ca2+], [Mg2+], and [TTC]tot were varied (details available in the Supporting Information). The light source used was designed to closely mimic the UV-A and UV-B portions of the terrestrial solar spectrum, and quantum yields for solar photolysis were determined for each significant species as described by Zepp (25).
Results and Discussion Speciation Studies. The fitting of pKa values to UV-vis spectral data gave the values 3.45 ( 0.10, 8.00 ( 0.05, and 9.78 ( 0.05 (at I ) 0), similar to those in the literature (3.30, 7.68, and 9.69; ref 26). The reported error in pKa values was calculated from adjustment of the parameters within the range in which no significant portion of the residuals was greater than the expected error of the absorbance measurements. These pKa values, although calculated from the total set of UV-vis spectra, are more easily visualized in the 400 nm cross-section of the titration data shown in Figure 2. The UV-vis absorptivity of TTC at 400 nm is shown as a function of pH, both with and without Ca2+ present. Calcium causes a decrease in the observed pKa values as well as an increase in the magnitude of light absorption, consistent with the binding of TTC to Ca2+ under the conditions listed. The subsequent fitting of the metal-binding association constants to the UV-vis absorption data yielded a simpler speciation scheme than that shown in Figure 1. The species L3-, ML-, M2L+, and M2HL2+ were not present at significant concentrations under the conditions investigated (pH ) 3-10, [Mg2+] ) 0-3 mM, [Ca2+] ) 0-32 mM, [TTC]tot ) 1525 µM). In other words, the fourth deprotonation did not occur to any significant degree, and there were no 2:1 metal: ligand complexes under the experimental conditions of this study. The observations and reasoning which led to this simplification of the speciation scheme are as follows. A minimum in RMS error of the spectral fit was first sought for the Ca2+/TTC UV-vis data, determined from the residuals of the total set of spectra as defined in the Supporting Information. Fitting of the data yielded a continuum of parameter space for which there was acceptably low error in reproduction of the experimental data. This continuum consisted of a range of possible KM2L and KML values, VOL. 40, NO. 23, 2006 / ENVIRONMENTAL SCIENCE & TECHNOLOGY
9
7237
TABLE 1. Equilibrium Constants for the Association of Ca2+ and Mg2+ with Aqueous TTC Species equilibrium
constant
log K
H2LHL2-
KCaH2L+ KCaHL KMgH2L+ KMgHL
3.4 ( 0.1a 5.8 ( 0.1 3.9 ( 0.1 4.1 ( 0.2
CaH2L+ CaHL0
+ h Ca2+ + h 2+ Mg + H2L h MgH2L+ Mg2+ + HL2- h MgHL0 Ca2+
a Error determined from the ability of the speciation model to reproduce experimental UV-vis absorbance measurements, except for KMgHL, for which it was determined log KMgHL < 4.3, and it was assumed KMgH2L+ < KMgHL.
throughout which KM2L > KML. This answer was rejected. It would require the association constant for the binding of a second metal with CaL- to be stronger than the first metal association with HL2-, which is unlikely. Second, because Ca2L+ would have been the major species throughout much of the experimental conditions range, the dissociation of the fourth acidic proton would be expected to be observable during a potentiometric titration containing [Ca2+] ) 10-1 M. This fourth proton was not observed (data not shown). These additional restrictions on the parameter space to be searched resulted in a new minimum corresponding to the Ca2+/TTC and Mg2+/TTC association constants listed in Table 1. The error values in Table 1 indicate the significance with which each binding constant contributed to the ability of the model to reproduce the spectral data, with the exception of KMgHL. Considering the limitations set by the range of conditions under which data were taken, it can only be said that log KMgHL < 4.3. This is because log KMgHL values greater than 4.3 require MgHL to be a significant species throughout a portion of the data for which no change in speciation was observable. The range of possible values for log KMgHL was narrowed further by the assumption that KMgHL > KMgH2L+, resulting in the reported value log KMgHL ) 4.1 ( 0.2. The TTC concentrations used in this study were kept below 25 µM because of concerns that TTC was associating with itself at higher concentrations. The bubbling of Ar gas through solutions containing [TTC] > 80 µM and any amount of Ca2+ or Mg2+ resulted in the formation of soapy bubbles. In addition to the inconvenience posed by the soapy characteristics at moderately high [TTC], this behavior was a clear indication of self-association, promoted by the metal ions. The decision to use 25 µM as a maximum TTC concentration was based on the limitations set by the sensitivity of UV-vis absorption measurements. Lower concentrations would have been preferable. It is suspected that due to the possibility of self-association, a systematic error greater than that reported may apply to some of the values reported here. Methods outside the scope of this study which may allow for investigations at lower TTC concentrations include the use of longer cell path length or fluorescence spectrometry. The observation that neither the loss of the fourth proton nor the binding of a second metal occur to any significant degree within the conditions investigated validates the association constants calculated by Martin (23). The binding constants found by Martin, determined through measurement of single wavelength UV-vis absorption as a function of metal ion concentration at constant pH (with 0.005 M Tris buffer present), were reported at an ionic strength of 0.15. Adjusting them to I ) 0 yields log KCaH2L+ ) 3.55, log KCaHL ) 4.98, log KMgH2L+ ) 4.01, and log KMgHL ) 5.21 (from the reported values 3.04 ( 0.02, 3.96 ( 0.05, 3.5 ( 0.03, and 4.19 ( 0.07, respectively, at I ) 0.15; see the Supporting Information for activity corrections). The equilibrium constants for the binding of Ca2+ or Mg2+ to the monoanion seem to agree well between Table 1 and ref 23. The binding constants with 7238
9
ENVIRONMENTAL SCIENCE & TECHNOLOGY / VOL. 40, NO. 23, 2006
respect to the dianion, however, disagree by approximately one log unit, and the order of strength is reversed. Martin reported log KMgHL > log KCaHL, whereas in the current study it was determined that log KCaHL > log KMgHL. There are many possible explanations for this discrepancy. The experiments by Martin require the assumption that a constant pH was maintained within each titration, which may have been difficult to accomplish. The tetracycline concentration was more than twice that used in the current study, offering a higher risk for the formation of multiple ligand complexes. Also, experiments involving titration via addition of a divalent metal are complicated by the inherent changes in ionic strength. UV-Vis Absorption Spectra. The fitting of binding constants to the UV-vis absorption data yielded the individual component spectra (calculations described in the Supporting Information). The UV-vis absorptivity spectra corresponding to each TTC species are shown in Figure 3. The species MgHL0 was not included in Figure 3 because the error in knowledge of its binding constant and the lack of experimental conditions which included a significant quantity of MgHL0 resulted in a range of possible spectra with a great deal of error. Knowledge of its spectrum, however, was not necessary for the purposes of this study, for it is not a significant species under environmentally relevant conditions. The speciation model and component spectra were tested using independent solutions containing both Ca2+ and Mg2+ as well as a filtered sample of Mississippi River water for which the metal ion concentrations were measured. Spectra were predicted within the expected margin of error. See the Supporting Information for details. The absorption feature of TTC most important to its environmental photochemistry is the peak at λ > 300 nm; this is the portion which overlaps with the solar spectrum. It is apparent from Figure 3 that this peak shifts to longer wavelengths as each consecutive proton is removed, and it increases in magnitude as TTC associates with Ca2+ or Mg2+. As a result, the rate of absorption of sunlight by TTC on a per-photon basis increases with either increased pH or metal ion concentration. Multiplication of the spectra in Figure 3 by the terrestrial solar spectrum leads to the following order of photochemical action (Ein mol-1 per unit time, with normalized magnitude): H4L+ (1.00) < H3L0 (1.12) < H2L(1.43) < CaH2L+ (1.53) < MgH2L+ (1.67) < HL2- (1.68) < CaHL0 (1.72). The change in photochemical action upon addition of metal ions is, in reality, more pronounced than simply the change in action between an anion and its respective metalbound complex. Changes in the macroscopic acid-base dissociation constants caused by metal-binding must also be taken into account. For example, binding of Ca2+ to the H2L- structure, to form CaH2L+, brings about a modest change in photochemical action (about 7%), but the concomitant increase in TTC acidity leads to formation of CaHL, a species with a more pronounced increase in photochemical action (20%). Photochemical Kinetics. Photolysis of TTC was observed to include some self-sensitization. An attempt to apply a pseudo-first-order rate law to the kinetic data spanning different initial TTC concentrations results in a higher apparent rate constant at higher initial [TTC], within the range of 2-20 µM. An example of this effect is shown in Figure 4. Because the kinetics are clearly not first-order, it would be misleading to label the y-axis of Figure 4 as a pseudofirst-order rate constant. It is labeled instead as a plot of the initial rate divided by the initial concentration (riC0-1) vs the initial concentration (C0). Extrapolation of these data by linear regression to C0 ) 0 is expected to provide a reasonable
FIGURE 3. Molar absorptivity of selected tetracycline component species: (a) acid-base forms and (b) metal-bound forms.
FIGURE 4. The initial rate of TTC photolysis divided by the initial concentration (-ri C0-1) vs the initial TTC concentration (C0) for the conditions pH ) 7.5, I ) 0.1. estimate of the direct photolysis pseudo-first-order rate constant, kdirect, at high dilution. Without the experimental variation of C0, the deviation from first-order kinetics may not have been noticed. Many individual kinetic time series appeared to be pseudo-firstorder (i.e. a plot of ln(C/C0) vs t appeared to be linear; see the Supporting Information), despite the fact that variation of C0 proved the process was not first-order in TTC under the given conditions. One possible explanation for the false appearance of first-order kinetics within an individual time series is that the production of reactive oxygen species (ROS) upon photolysis of TTC occurs concurrently with direct photolysis. Total loss of TTC is then the sum of loss by direct photolysis and reaction with ROS. In such a kinetic scheme, as the initial [TTC] approaches infinite dilution, the observed loss should approach that of pseudo-first-order direct photolysis in the absence of self-sensitization. For the appearance of near first-order loss observed in this study to have occurred, one or more of the products of photolysis must have photosensitizing capabilities similar to TTC. This is in agreement with the findings of the study by Miskoski et al. (12), in which it was observed that one of the products of TTC photolysis has a higher quantum yield for the production of singlet oxygen than that of TTC. Although this mechanism implies that TTC photolysis in surface waters may be enhanced by naturally produced ROS, the direct photolysis of TTC is rapid enough such that reaction with environmental ROS would be insignificant. Considering average singlet oxygen concentrations of less than
3 × 10-14 M at the surface, under noon, summer sunlight (27), it is unlikely to be able to compete with a direct photolysis half-life on the order of 1 h. Another possible explanation for the observed selfsensitization effect is that TTC associates with itself even below 20 µM. This, however, cannot be a complete explanation. In the absence of metal ions, the magnitude of the selfsensitization effect increases with each proton removed. It is greatest under the conditions where the dianion HL2dominates. It seems unlikely that the dianion would associate with itself in the absence of metal cations. Self-sensitization for the loss of species HL2- was, in fact, so much greater than loss due to direct photolysis, that a kdirect for HL2- could not be determined. Some self-sensitization was also observed in the presence of Ca2+ and Mg2+, though the magnitude was much less. The magnitude of self-sensitization, here, refers to the slope of a plot analogous to that in Figure 4 relative to the direct photolysis rate constant. The kinetic data were extrapolated back to infinite dilution of TTC. The series of photolysis experiments performed and the resulting kdirect values resulting from extrapolation to C0 ) 0 are listed in Table 2. Quantum yields for the solar photolysis of each TTC form are also included in Table 2. The error indicated for each rate constant value was calculated from the error in determination of the intercept of the plot of riC0-1 vs C0. Error in quantum yield values were propagated from the uncertainty in the kdirect values and weighted appropriately corresponding to their contribution to the knowledge of the given quantum yield. After extrapolation of the kinetic data to infinite dilution, the direct photolysis quantum yield value for HL2- can, at best, be reported to be
