Effect of Water Quality on Rejection of Selected Human and Veterinary Antibiotics by Nanofiltration and Reverse Osmosis Membranes

This study addresses the possibility for nanofiltration (NF) and reverse osmosis (RO) membrane processes to serve as an alternative to remove low levels of selected antibiotics from wastewaters and surface waters given changing water conditions such as pH and presence of natural organic matter (NOM). Preliminary experimental results obtained with a dead-end filtration apparatus suggest that adsorption of antibiotics onto the membrane surface is negligible. However, removal is greatly affected by changes in pH due to changes in electrostatic repulsion between charged antibiotics and membrane surfaces. Both the pore diffusion model for NF and the solution diffusion model for RO will be applied to further understand the mechanisms and trends behind the experimental data. INTRODUCTION In recent years, antibiotics have emerged as a group of contaminants with particular concerns because of their frequent detection at low levels (μg/L) in wastewaters and surface waters in the US, Canada, and parts of Europe 2, . Widely used for both human therapy and veterinary practices, antibiotics are only partially consumed by the users and can enter the environment through wastewaters and animal manure. Researchers indicate that such recalcitrance of these compounds at low levels may promote bacterial resistance and cause adverse human health effects 5, . Previous studies have shown that certain water treatment processes can affect the fate and transformation of antibiotics. Disinfectants such as free chlorine, chloramine, chlorine dioxide and ozone are proven to effectively oxidize various antibiotics 8, , but the transformation of parent compounds into a more environmentally potent species is of growing concern. Pressuredriven nanofiltration (NF) and reverse osmosis (RO) membrane processes are emerging technologies which may be an important alternative towards removing antibiotics from wastewaters and surface waters. Both NF and RO have been suggested as an effective barrier against organic compounds such as pesticides and steroid hormones under various water quality conditions . However, few studies have been reported on antibiotics removal by membrane processes to date. Considering their low molecular weight and charged speciation, removal of antibiotics might pose a unique challenge to water and wastewater treatment utilities. The purpose of this study is to evaluate the potential removal of selected human and veterinary antibiotics, such as sulfamethoxazole (SMX), trimethoprim (TMP), erythromycin (ERY), ciprofloxacin (CF), and carbadox (CDX), by NF and RO membranes under varying water quality parameters and antibiotic concentrations, with the goal of understanding the underlying rejection mechanism(s). These antibiotics are selected considering their persistent use in the environment as well as variable pH-dependent charge speciation. As water quality parameters such as pH and the presence of NOM change, the effects of membrane surface charge, antibiotic concentration and antibiotic speciation on overall rejection and/or adsorption will be assessed. Both pore diffusion and solution diffusion models will be applied to experimental data to further elucidate antibiotic-membrane interactions. EXPERIMENTAL METHODS Materials. Five commercial NF and RO membranes were selected from two manufacturers, Saehan Industries Inc. and Filmtec (Dow Chemical Company). Saehan provided two NF membranes, NE 4040-90, NE 1812-70, and one RO membranes, RE 4040-BLN. Filmtec provided two NF membranes NF 270-400 and NF 90-400. All membranes are composed of a polysulfone support with a polyamide active layer but vary in divalent and monovalent ion rejection levels. Chemicals and Solution Preparation. Ciprofloxacin hydrochloride and sulfamethoxazole were purchased from ICN Biomedicals, Inc. (Irvine, CA, USA). Trimethoprim, carbadox, and erythromycin were purchased from Sigma-Aldrich Co. (St. Louis, MO, USA). Structure and relevant properties of the antibiotics used in this study are shown in Table 1. All the chemicals used were reagent or analytical grade, and all the synthetic solutions were prepared using Nanopure water produced from a Millipore water purification system. Antibiotic stock solutions were prepared at 25 or 100 mg/L in 10 to 50% methanol. Stock buffer solutions consisted of acetate (pH 4 -5), phosphate (pH 6-8), and borate (pH 9). Stock solutions were then spiked into test solutions to achieve a final antibiotic concentration of 100 μg/L or lower to represent both elevated and environmentally relevant concentrations. The final buffer concentration of test solutions is 1.0 mM. Solution pH is adjusted by addition of H2SO4 or NaOH. Membrane Preparation and Performance Testing. Flat sheet membrane coupons were cut and soaked in Nanopure water for 24 hours prior to testing. To ensure comparable membrane quality, salt rejection was tested and compared to the manufacturer’s specifications. Only membranes that passed the initial quality tests were used. For NF membranes, rejection was measured using a feed solution of 2,000 mg/L MgSO4 at 70 or 75 psi. All the NF membranes passing the quality testing fell within + 3 % of specified MgSO4 rejection of 97% or greater. For RO membranes, rejection was measured using a feed solution of 1,500 mg/L NaCl at 150 psi. All the RO membranes passing quality testing fell within +3% of specified NaCl rejection of 99%. Dead-End Cell Experiments. All experiments were performed using a stainless steel dead-end stirred cell (300 ml volume, 14.6 cm active membrane area). Water permeation flux was measured both when using Nanopure water for 30 min prior to the experiments as well as between sampling events during test runs. Comparable fluxes for all membranes were achieved by altering operating pressure. Membrane coupons were used repeatedly for multiple experiments. Antibiotics were desorbed from the membrane between experiments by running 150 mL of Nanopure water at 140 psi. Complete desorption was confirmed by comparing adsorption trends of replicate experiments. All experiments were performed at room temperature. Efforts were made to maintain antibiotic concentration in the retentate as constant as possible by returning collected permeate to the feed every 10 min by depressurizing and opening the stirred cell for an overall run time of 60 to 100 min. Permeate samples (sample volume of ~ 0.5 ml) were collected repeatedly at times less than 10 minutes in order to evaluate adsorption effects. However, samples collected after 10 min were taken immediately following each permeate recirculation event in order to neglect effects of increasing retentate concentration. All the permeate samples were collected in 1.5 ml amber HPLC vials except for ciprofloxacin samples which were collected in 20 ml borosilicate amber vials and further acidified with 1.0 M H3PO4 before analysis. The experimental design of all dead-end experiments followed a matrix where parameters such as pH, NOM concentration, and antibiotic concentration were varied. pH effects on antibiotic rejection were studied in NF by varying pH from 0.5 to 2.0 units away from environmentally relevant pKa values. In on-going study, NOM are added until when competition with significant levels of antibiotic adsorption is possible. Analytical Methods. Concentrations of all antibiotics were measured using an Agilent 1100 Series HPLC system equipped with an RX-C18 reverse phase column. Concentrations of SMX, TMP, and CDX were measured using a UV/VIS diode-array detector while CF was measured using a fluorescence detector. Method development for CF, SMX, TMP, and CDX was taken from prior studies by using gradients of buffer solutions, 0.04 mM H3PO4 (CF), 1.0 mM TFA (SMX and TMP), or 20 mM acetic acid (CDX), and acetonitrile. A method for ERY has not yet been developed. To assess antibiotic concentrations of a few μg/L, samples will be preconcentrated using solid phase extraction (SPE) prior to HPLC analysis. Structure MW pKa Charge at pH 7 Ciprofloxacin 331.4 pKa,1=6.2 pKa,2 = 8.8 Zwitterion and Neutral Sulfamethoxazole 253.28 pKa,1= 1.7 pKa,2 = 5.6 Anion Trimethoprim 290.3 pKa,1=1.32 15 pKa,2 = 7.45 Cation and Neutral Erythromycin 733.9 pKa,1=8.90 Cation Carbadox 262.2 No pKa found between pH 3-11 N/A Table 1. Structures and Properties of Antibiotics Used in This Study