Introduction

A variety of contaminants in municipal sewage pose potential risks to human health and aquatic ecosystems (1, 2). For example, pathogens present in untreated sewage have caused widespread outbreaks of gastrointestinal illness (3) while chlorine-disinfected wastewater effluent is believed to be the most important source of the carcinogenic disinfection byproduct N-nitrosodimethylamine (NDMA) and its precur sors in locations where water reuse is practiced (4). In surface waters where wastewater effluent accounts for a significant fraction of the overall flow, fish may be impacted by steroid hormones that interfere with development and possibly reproduction (5, 6). A variety of other wastewater-derived chemical contaminants (WWDCs) have also been detected in surface waters (7, 8), but no known risks associated with exposure of humans (9) or aquatic organisms to low concentrations of these compounds have been identified.

Most studies of fate and transport of WWDCs treat wastewater effluent as the sole source of wastewater-derived contaminants in surface waters and use the simplifying assumption that wastewater treatment plant (WWTP) performance is equivalent during wet and dry weather (10, 11). Under these conditions, the concentration of WWDCs in surface waters should be highest in systems where effluent is the main source of water. Although wastewater effluent is undoubtedly an important source of contaminants in many locations, wet weather flows may also contribute WWDCs to surface waters. For example, higher concentrations of some WWDCs have been observed during wet weather than under base flow conditions in a lake in Switzerland (12) and in several U.S. rivers (13, 14).

In the U.S., there are approximately 800 cities with a total population of 40 million people that have combined sewers (15). Most of these systems discharge raw sewage into receiving waters during rainfall or snowmelt events through combined sewer overflows (CSOs). Direct discharges of raw sewage also can occur during dry weather through leaking sewers and improperly functioning septic tanks. Although the higher flows in surface waters normally associated with CSOs should dilute WWDCs, CSOs can still be an important source because the concentrations of WWDCs in raw sewage can be up to 1000 times greater than those present in wastewater effluent (12, 16, 17). Furthermore, most WWTPs remove WWDCs less effectively during wet weather events (17). As a result, relatively high concentrations of some WWDCs may be present in surface waters during high-flow conditions.

Insight into the relative contribution of treated and untreated sewage to WWDCs in surface waters might be gained by comparing the concentrations of recalcitrant compounds and compounds that are removed efficiently by WWTPs. However, this would not be a very accurate means of quantifying contributions of untreated sewage because concentrations in raw and treated sewage can vary widely over short periods of time (6, 17, 18). Additionally, this type of analysis would be complicated by removal of the labile WWDCs via attenuation in surface waters.

An alternative approach for quantifying the relative contribution of raw sewage to surface waters involves the measurement of enantiomers of chiral WWDCs. The chirality of certain pesticides has been used previously by researchers as a means of apportioning their sources (19, 20). The relative concentration of a compound's enantiomers is often expressed as the enantiomeric fraction (EF), which is defined by eq 1

In this study we examined the use of propranolol, a chiral pharmaceutical, as a tracer of raw and treated sewage. Propranolol was selected for this purpose because its enantiomers can be separated after derivitization by gas chromatography without an enantioselective stationary phase. Propranolol is a -blocker that is one of the most common prescription drugs in the United States (21). It has been detected in European surface waters at concentrations up to 590 ng/L and in European and American wastewater effluents at concentrations up to 290 and 1900 ng/L, respectively (17, 22). To assess its utility as a tracer, samples were collected from WWTPs and surface waters where previous research had shown that wastewater effluent or raw sewage are important sources of WWDCs. These measurements were complemented by laboratory studies with microcosms designed to simulate WWTPs and effluent-receiving surface waters.

Experimental Section

Materials. Racemic and enantiomerically pure propranolol hydrochloride (99%), metoprolol (99%), N-methyl-N-trifluoroacetamide (MSTFA) hexachlorocyclobenzene (HCB) (99%), and (-)-()-methoxy--(trifluoromethyl)phenylacetyl chloride ((-)-MTPA-Cl) were obtained from Sigma Aldrich (St. Louis, MO). Analytical grade methanol and isooctane, sodium chloride (NaCl), tryptic soy broth, as well as sodium azide were obtained from Fisher Scientific (Pittsburgh, PA). Distilled water treated with a Barnstead Nanopure II system was used for sample preparation.

Sampling Locations. Raw sewage and final effluent samples were collected from seven WWTPs in California and New York (Table 1). At four of the WWTPs, samples were collected at different stages of the treatment train. The wastewater collection system for the 26th Ward WWTP is a combined sewer, and the WWTP has a bypass system for wet weather. This WWTP was sampled during dry weather when the bypass system was not being used and during wet weather when sewage that bypassed secondary treatment comprised 9% of the flow. In addition, sewage was sampled from the wastewater collection system near Bush Creek at Ward Street in Kansas City, MO.

All wastewater was collected as grab samples (with the exception of the samples from the San Jose/Santa Clara Water Pollution Control Plant, which were 24-h flow-weighted composites) in 4-L amber glass bottles spiked with 8 g of NaCl to prevent sorption of the analyte onto the sides of the vessel.

Grab samples also were collected from four surface waters in 1- or 4-L amber glass bottles spiked with 2 g/L NaCl. Each surface water sampling site is described below.

Samples were collected from the Mt. View Sanitary District's engineered treatment wetland, in Martinez, CA, which was constructed to supply additional treatment to the effluent of the adjacent WWTP. The wetland consists of five ponds in series connected by weirs and underground pipes. Each pond is between 0.4 and 1.2 m deep and is vegetated around the edges with cattails and bulrushes. Results of a LiCl tracer test indicated that the hydraulic retention time of the wetland is approximately 9 days. Samples were collected at the inlet of the wetland, at a mid-wetland location where the water was determined to have a hydraulic retention time of approximately 5 days relative to the wetland inlet, and at the wetland outlet.

Samples were collected upstream and downstream of a storm sewer outfall into Gwynns Falls, Baltimore County, MD. The storm sewer is known to be contaminated by raw sewage from a nearby leaky sanitary sewer. Gwynns Falls otherwise does not receive any wastewater effluent discharges (23).

Samples also were collected from Jamaica Bay (Figure 1), which is located on the south shore of western Long Island, NY. During dry weather, the bay receives approximately 12.9 m³/s (290 million gallons per day (MGD)) of secondary effluent. During wet weather, raw sewage and stormwater runoff are discharged into the bay from CSOs. The bay is on average 5 m deep, and the water in the bay has a hydraulic residence time of approximately 35 days (24). Samples were collected from Jamaica Bay off the nearby Canarsie Pier in Brooklyn, NY, 1 and 4 days after a rainfall event. This site is impacted by treated wastewater effluent from the 26th Ward WWTP as well as the CSO outfalls of both this plant and the Coney Island WWTP. A sample also was collected from a lagoon that receives CSO runoff from near the 26th Ward WWTP. The lagoon, which is located adjacent to the 26th Ward WWTP, drains into Jamaica Bay.

Figure 1 Sampling sites in Jamaica Bay, NY.

Finally, samples were collected from the Santa Ana River, which is located in Orange and Riverside Counties, CA (Figure 2). The Santa Ana River flows approximately 160 km from the coastal mountains near Big Bear Lake to the Pacific Ocean. During summer, most of the flow of the river is attributable to twelve WWTPs that discharge into the Santa Ana River and its tributaries (25, 26). The three largest WWTPs that discharge into the Santa Ana River Basin are Rapid Infiltra tion/Extraction (RIX), Riverside WQCP, and Chino RPI. Up to 50% of the river's flow is diverted into the Prado Wetlands, which are located approximately 50 km upstream of the Pacific Ocean. Upstream of the Prado Wetlands, the river is approximately 0.5 m deep and has a sandy bottom whereas downstream of the wetlands the river is confined to a concrete-lined channel as it passes through an urban area.

Figure 2 Sampling sites along the Santa Ana River, CA. WWTPs are represented by stars. Letters indicate locations where surface water samples were collected.

Samples were collected from various reaches of the Santa Ana River on three occasions. During the first two rounds, grab samples were collected from different reaches of the river, and one sample from Chino Creek, which is a tributary to the Santa Ana River, during a period of 12 h. The third round was conducted as part of a synoptic study, following three parcels of water down an 11-km stretch of the Santa Ana River between the RIX facility and the Riverside WQCP.

Samples from surface waters and wastewater treatment plants were immediately placed in coolers with ice for transport back to the laboratory, where they were stored at 4 C. The samples were filtered through 0.45-m glass-fiber filters within 4 days of sampling.

Microcosm Experiments. To simulate the biotransformation of propranolol in a municipal WWTP, five 4-L amber glass bottles were filled with 3.9 L of filtered secondary effluent from the East Bay Municipal Utilities District's WWTP, collected prior to disinfection. Three of the treatments were amended with 20 mL of freshly collected return activated sludge containing 3000 mg/L suspended solids from the same WWTP. One of the three bottles containing activated sludge was sterilized with 40 mM sodium azide. One of the two treatments that did not receive return activated sludge was sterilized with 10 mM sodium azide. Before the experiment was started, 10 mg/L tryptic soy broth was added to each of the bottles that received return activated sludge. An additional 10 mg/L was added to the two bottles with activated sludge that had not been sterilized on each subsequent day of the experiment.

Humidified air was bubbled through each of the bottles through a glass frit at a rate of 150 mL/min. The system was allowed to equilibrate for 24 h prior to addition of 1000 ng/L racemic propranolol. Approximately 500-mL samples were collected daily over a period of 6 days. Although this is longer than the hydraulic retention time, a longer period was needed due to differences between the microbial processes in the batch microcosm and the full-scale system. Before extraction, each sample was fortified with 0.5 g of racemic metoprolol as an internal standard, which resulted in a concentration of metoprolol at least an order of magnitude higher than that which was already present in the matrix. Suspended solids and pH were monitored over the course of the experiment using standard methods (27).

To simulate biotransformation of propranolol in surface waters, six 4-L clear Pyrex beakers were filled with water from the Mt. View Wetland that had been strained through a 75-m sieve. The initial concentration of propranolol in the wetland water was less than 10 ng/L. The beakers were spiked with 1000 ng/L of racemic propranolol and placed in a 15 C constant-temperature bath, located on the roof of a building on the University of California at Berkeley campus that receives exposure to direct sunlight. Three of the beakers were covered in aluminum foil. One of the beakers covered in foil and one beaker exposed to sunlight were sterilized with 10 mg/L sodium azide. Aliquots with volumes of 200, 300, 400, 500, and 1000 mL were removed for analysis on day 0, 2, 7, 13, and 20, respectively. Before extraction, each of these samples was fortified with 1 g of metoprolol as an internal standard, which resulted in a concentration of metoprolol at least an order of magnitude higher than that which was already present in the matrix. To maintain constant sunlight exposure conditions, each time a sample of water was removed for analysis, it was replaced with the same volume of strained wetland water that had been stored at 5 C. Normalized concentrations of propranolol were calculated by accounting for dilution that occurred when water was replaced.

Analysis. Samples were subjected to solid-phase extrac tion prior to derivatization and gas chromatography/tandem mass spectrometry (GC/MS/MS) analysis. Prior to use, cartridges containing 500 mg of Supelclean C₁₈ resin (Supelco, St. Louis, MO) were conditioned with 10 mL of methanol followed by 20 mL of Nanopure water. Aliquots of between 0.5 and 2 L of sample were passed through the cartridges at a rate of 20 mL/min, followed by a rinse with 50 mL of Nanopure water. After extraction, the analytes were eluted from the cartridges with 10 mL of methanol. The eluent was dried overnight in a vacuum oven at 30 C, redissolved in 2 mL of methanol, transferred to 4-mL screw-cap vials, and blown to dryness under a gentle stream of purified nitrogen.

Propranolol and metoproplol were derivitized with MSTFA and (-)-MPTA-Cl using a method adapted from Kim et al. (28). After 50 L of MSTFA was added to the reaction vials containing the dry samples, they were capped and held at room temperature for 30 min, then placed in a 60 C oven for 5 min. After the samples cooled, 5 L of (-)-MPTA-Cl was added to the vials, which were then capped and returned to the 60 C oven for an additional 5 min. The derivitized samples were diluted with 100, 200, or 250 L isooctane that contained 500 g/L HCB as an internal standard to correct for variability in the injection volume.

Propranolol and metoprolol derivatives as well as HCB were analyzed by GC/MS/MS (Thermoquest, San Jose, CA). A 30-m, 0.25-mm (i.d.), 0.25-mm (film thickness) MDN-5S column (Supelco, Bellefonte, PA) was used for separation. Splitless injections of 1 L into an injection port set to 250 C were used. Helium was used as the carrier gas at 1.0 mL/min. The programmed temperature run consisted of an initial 1.0-min hold at 100 C, followed by a 3 C/min ramp to 300 C with a 1.0-min hold at the end of the program. Mass spectrometer conditions included electron-impact ionization at 70 eV in a 200 C ion source with a 300 C transfer line from the gas chromatograph. Details of the retention times and mass spectrometer conditions are listed in Table 2.

The limit of detection of propranolol using this method ranged from 0.1 to 1.0 ng/L depending upon the sample matrix and the instrument response. Spike recoveries of 150 ng/L of propranolol in Nanopure water with 2 g/L NaCl varied from 15% to 90% relative to HCB with a median of 71%. Recoveries were better when using freshly silanized glassware for sample processing. Concentration data were not corrected for recoveries since recoveries varied between samples within each set of samples. Due to these variations in recoveries within each batch of samples, we estimate an uncertainty in the concentration values we report of approximately ±30%.

The order of elution of the two enantiomers of propranolol from the GC column was determined by comparing the retention times of the sample with that of a standard solution of enantiomerically pure (S)-(-)-propranolol. In this paper, we report the EF of propranolol as the ratio of the concentration of the (R)-(+) enantiomer to the total concentration of propranolol.

Results

Chiral Analysis. The derivitization procedure for propranolol presented here was adapted for environmental analysis from a method developed for biomedical applications (28). In this method, the two enantiomers of propranolol were derivatized with a single enantiomer of a chiral compound, (-)-MPTA-Cl, to form diastereomers (Figure 3). The two propranolol-derivative diastereomers had different physical properties from one another and therefore could be separated with baseline resolution using a common reverse-phase gas chromatographic column (Figure 4). By comparison of the chromatographic retention time of a standard of enantiomerically pure (S)-(-)-propranolol to that of racemic propranolol, it was determined that the first peak to elute from the GC column corresponds to (R)-(+)-propranolol and the second to (S)-(-)-propranolol.

	Figure 3 Derivitization of (R)-(+)-propranolol and (S)-(-)-propranolol by MSTFA and (-)-MPTA-Cl.
	Figure 4 Chromatogram of a racemic 600 ng/L standard of propranolol. The peak labeled (1) is (R )-(+)-propranolol; (2) is (S)-(-)-propranolol.

Analysis of approximately 20 duplicate samples showed good agreement with a standard deviation for the EF of less than 0.03 whenever the concentration of propranolol was greater than 1 ng/L. The standard deviation of the EF was not correlated with the magnitude of the EF. There also was not a relationship between EF and concentration (Figure 5). Measurements of EF were not affected by sample-to-sample variation in extraction efficiency or derivitization yield.

Figure 5 Relationship between EF and concentration in samples taken at WWTPs.

In samples from WWTPs, the chromatogram of derivitized propranolol was accompanied by two peaks with retention times of approximately 1 min longer than propranolol (Figure 6). This second compound did not interfere with the measurement of propranolol in any of the samples.

Figure 6 Chromatogram of propranolol from Mt. View Wetland inlet. The peak labeled (1) represents the derivative of (R )-(+)-propranolol; (2) is the derivative (S)-(-)-propranolol; (3) and (4) are a doublet of unknown identity that consistently shows up in chromatograms of wastewater. In this sample, the total concentration of propranolol is 160 ng/L and EF = 0.37.

Concentrations of propranolol at the WWTPs varied from 13 to 250 ng/L in the plant influent and from 3 to 160 ng/L after biological treatment (Table 3). In individual plants where both the influent and the effluent were sampled, the reduction in the concentration of propranolol varied from 15% to 95% with a median of 77%. Although these removals are imprecise due to the use of grab samples and variation in recoveries, the data agree with previous studies that show that propranolol is partially removed during biological wastewater treatment (17). In each of the wastewater treatment plants surveyed, the propranolol was racemic in the influent (i.e., EF = 0.49-0.54) but not in the effluent (i.e., EF = 0.31-0.44). A paired Student's t-test of the five data sets where both influent and effluent were sampled showed that the EFs were significantly different in the influent and effluent (p < 0.014).

The wastewater effluent at the 26th Ward WWTP during dry weather had an EF of 0.42 for propranolol, which was the highest of any of the WWTPs. The high EF was consistent with the apparent poor removal during secondary treatment. During wet weather, when 9% of the effluent sample consisted of raw sewage that bypassed secondary treatment, EF of the effluent increased to 0.44.

At the San Jose Water Pollution Control Plant and at Mt. View Sanitary District, samples were taken at intermediate stages in the treatment process (Figure 7). In these samples EF decreased after every step of biological treatment and remained unchanged after chemical or physical treatment steps (i.e., filtration, settling, and chlorination).

Figure 7 Concentration and enantiomeric fraction of propranolol at different steps in the treatment train at (a) Mt. View Sanitary District, Martinez, CA, and (b) San Jose Wastewater pollution control district. For part a, the secondary treatment is a trickling filter and tertiary treatment is nitrification. The final treatment steps are filtration and UV disinfection. For part b, secondary treatment is activated sludge, which is followed by nitrification, filtration, and chlorination. Where duplicate samples were measured, vertical lines indicate the range of values.

In the microcosms designed to simulate an activated sludge tank, the concentration of propranolol decreased in the two treatments containing live activated sludge but did not change in either of the sterilized treatments or the treatment containing only filtered effluent (Figure 8a). The EF of propranolol decreased from racemic to 0.44 and 0.43 in the two activated sludge treatments after 6 days of incubation (Figure 8b).

Figure 8 The (a) normalized concentration and (b) EF values of propranolol in activated sludge microcosms and controls.

At the Mt. View Wetland, the concentration of propranolol in the wetland decreased from 230 to 94 ng/L between the wetland inlet and the outlet while the EF remained between 0.32 and 0.39 with no evident trend as the water passed through the wetland (Figure 9). While this decrease in concentration may have been attributable to photolysis or biodegradation, it is also possible that the apparent removal of propranolol was due to fluctuations in concentrations discharged by the WWTP.

Figure 9 Concentration and EF of propranolol in Mt. View Wetland. Vertical lines indicate the range of values for duplicate measure ments.

The behavior of propranolol in the Mt. View Wetland was consistent with the surface water microcosm experiment in which the concentration of propranolol decreased in each of the microcosms except the treatment that was sterilized and protected from sunlight (Figure 10). The concentration of propranolol in the microcosm that was sterilized and exposed to sunlight decreased with a first-order rate of 0.13 ± 0.02 day^-1, indicating that phototransformation took place. The concentration of propranolol decreased at a rate of 0.08 ± 0.02 day^-1 in the microcosms that were protected from the sunlight but not sterilized, indicating that biotransformation also occurred. The concentration of propranolol in the microcosm that was exposed to sunlight and not sterilized decreased at a rate of 0.17 ± 0.04 day^-1. There was no decrease in concentration in the dark, sterilized micocosm, indicating that sorption to fine particles or to the walls of the container did not take place. The propranolol remained racemic (i.e., EF = 0.48-0.52) in all 30 samples collected from the six microcosms over the 20 days of the experiment.

Figure 10 Normalized concentration of propranolol in surface water microcosms and controls.

There was significant scatter in the concentration data for both microcosms. Although this may have been attribut able to inconsistencies in the rate of degradation throughout the course of these experiments, it also may have been due to the use of metoprolol as a surrogate standard. Although normalizing measured concentrations of propranolol against known concentrations of metoprolol yields better results than not using corrected recoveries, the recovery of metoprolol often differs from that of propranolol.

At the two surface water sites where raw sewage was expected to be important, propranolol detected in surface waters was racemic. Propranolol was detected at a concentration of 32 ng/L downstream of the contaminated storm sewer in Gwynns Falls but not upstream. Propranolol detected below the storm sewer was racemic (Table 4). Propranolol detected in Jamaica Bay on May 26, 2004, and in the CSO lagoon adjacent to the 26th Ward WWTP on April 27, 2004, was close to racemic (Table 4).

In the first reach of the Santa Ana River, between the RIX WWTP and the Riverside WQCP (i.e., sites i-v), the EF values were comparable to those of wastewater effluent. The concentration of propranolol was below the limit of quantification in the sample collected in June 2004 and in all 15 samples from the synoptic study conducted in September 2004. In the three instances where it was detected at concentrations below the limit of quantification, the EF ranged from 0.21 to 0.41. In the downstream sections of the Santa Ana River and in Chino Creek, propranolol was detected in June 2003, in six of the eight samples at concentrations up to 31 ng/L. The EF of propranolol measured at each of these sites (i.e., 0.42-0.53) was comparable to the values that would be expected in a mixture of wastewater effluent and raw sewage (Table 4).

Discussion

The EF of propranolol decreased from racemic to values significantly below racemic during biological wastewater treatment in each of the seven WWTPs that were sampled in this study. The enantioselective degradation of propranolol also was observed in the simulated activated sludge system (Figure 8b). In contrast, the EF remained unchanged as water passed through the Mt. View Wetland and during the 20-day incubation in the surface water microcosm experiment.

These results suggest that when propranolol is detected in surface waters an EF of 0.50 ± 0.03 indicates that its source is raw sewage and an EF of 0.42 or less indicates that its source is treated sewage. An EF between these values likely indicates contributions from multiple sources. While quantitative source apportionment may be possible in some systems, the EF of propranolol emerging from WWTPs can be somewhat variable (Table 3). Care should be taken in source apportionment due to the uncertainty associated with the EF from a given WWTP.

Although it is possible that the wetland and the surface water microcosm described here are not representative of all surface water systems where biotransformation could occur, the rate of biotransformation of propranolol in surface waters is likely to be slow relative to the hydraulic residence time of most effluent-impacted rivers. In other words, the EF of propranolol would probably not change significantly after discharge into a river because the hydraulic retention time of most effluent-impacted rivers is less than a few days, which is considerably slower than rates of propranolol biotransformation expected in surface waters. It is also possible, but unlikely, that propranolol undergoes enantiomerization (the transformation of one isomer to the other) in surface waters. For this to be the case and for the EF to remain constant in surface waters, the rate of enantiomerization would have to equal the difference in rates of biodegradation between the two enantiomers. This would not affect the utility of propranolol as a tracer. This pattern of enantioselective degradation by sewage sludge and the absence of enantioselective degradation in surface waters has been observed previously for the chiral pesticide metolachlor (29, 30).

Pharmacological studies have shown that propranolol, which is administered as a racemate, is absorbed into the intestine in a manner that is not enantioselective, with 10% being excreted directly through feces. Once in the bloodstream, it undergoes glucuronidation, side-chain oxidation and ring oxidation (31). Of the initial dose, 15% is excreted in urine as a glucuronide-propranolol conjugate with an EF of approximately 0.41, with some variability between individuals (32, 33). Neglecting sources of propranolol unrelateted to human feces and urine (e.g., disposal of unused product), we would expect the EF in WWTP influent to be 0.45 on the basis of a two-source apportionment model (34). Because the propranolol that we observed in WWTP influent and a sewer line were racemic, we hypothesize that its source was feces, and the propranolol glucuronide excreted in urine remains largely intact while being transported through the sewers. Once in the WWTP, the glucuronide likely was cleaved by glucuronidase enzymes present in bacteria (35). If this is the case, then only part of the enantioselective degradation observed in WWTPs was attributable to biological treatment-some may be due to the cleavage of nonracemic propranolol glucuronide.

Although there are numerous sources of propranolol in Jamaica Bay that are difficult to quantify, our data are consistent with the hypothesis that much of the propranolol in Jamaica Bay originates from CSOs and bypass flow during rainfall events. The EF of propranolol detected in Jamaica Bay was 0.48 on May 26, 2004, 1 day after a rainfall event of 0.5 cm, whereas the concentration of propranolol was below the limit of detection on September 13, 2004, 4 days after a rainfall event of 7.8 cm.

At the 26th Ward WWTP, the EF of the propranolol in the effluent was 0.42 during dry weather, and it was 0.44 during the wet weather period when 9% of wastewater by volume bypassed secondary treatment. These EF values are consistent with a two-source apportionment model (19), where raw sewage and secondary effluent are the two sources, even though the 26th Ward WWTP removes propranolol poorly compared to the other WWTPs surveyed. Additionally, previous studies have shown that the efficiency of removal of pharmaceuticals from WWTPs is adversely affected by wet weather (17), so the EF of propranolol in the secondary effluent during wet weather could be higher than at the same plant during dry weather.

At Gwynns Falls, a fairly simple system where a leaky sewer was known to exist, the EF downstream of the storm sewer discharge point was 0.50, a value characteristic of untreated sewage. Therefore the EF value correctly identified the source of WWDCs in the river.

In the Santa Ana River, in the three samples in which propranolol was detected between RIX and Riverside WQCP, the compound was not racemic, indicating that the major source of propranolol in this reach was the RIX plant and the smaller WWTPs upstream. However, in Chino Creek and the sections of the river downstream of the Prado Wetlands, propranolol had an EF that was characteristic of untreated sewage or a mixture of treated and untreated sewage. Although the WWTPs that discharge into Chino Creek were not sampled, we assume that they function as well as the other WWTPs in the area and discharge low concentrations of nonracemic propranolol. Therefore, our data suggest that a significant fraction of the propranolol in this section of the Santa Ana River is related to raw sewage discharges.

The measurement of WWDCs in surface waters can be a useful alternative to indicator organisms for the detection of leaky sewers. Methods using biological indicators of sewage are subject to false positives due to the presence of similar organisms derived from wildlife or other sources unrelated to human waste (36, 37). The measurement of WWDCs is also free of the many complications that confound accurate quantification and source tracking of bacteria and viruses in environmental samples (36, 37). The use of the kind of enantiomeric chemical analysis presented here has the further advantage that it can distinguish raw and treated sewage and therefore could be applied to detect raw sewage in effluent-impacted waterways.

Three decades ago, it was estimated that in the U.S. 7.7 million people were served by drinking water utilities using in excess of 50% wastewater as their raw water source during low river flow conditions (38). It is likely that number has since increased along with the elevated demand for water that accompanied the population growth in arid regions. Accordingly, concerns about adverse health effects due to exposure to WWDCs in drinking water derived from effluent-dominated rivers are increasing. Tools such as the Pharmaceutical Assessment and Transport Evaluation (PhATE) model (10) have been designed to predict concentrations of WWDCs in surface waters used as drinking water sources. These models may underestimate concentrations at drinking water intakes because they do not account for inputs of WWDCs from sewage derived from CSOs, leaky sewers, and WWTP bypass flows. In particular, concentrations of WWDCs that are removed effectively in WWTPs may be underestimated. For example, caffeine and ibuprofen are both compounds present at high concentrations in raw sewage, and removals above 99% typically occur during wastewater treatment (11, 16). For such compounds, raw sewage may be responsible for the majority of their loading into surface waters (12, 16).

The method that we have described here is useful for determining whether a waterway is significantly impacted by untreated sewage. It can be used qualitatively to apportion the contributions of treated and untreated sewage into surface waters for water quality and fate and transport studies and to help inform models predicting the concentration of WWDCs in drinking water. Additionally, this kind of enantiomeric analysis could be used in conjunction with microbial source tracking as a means of ascertaining the contribution of leaking sewers to pathogens in a waterway.

Acknowledgment

The authors thank Upal Ghosh (University of Maryland, Baltimore County, MD), Bruce Brownawell (Stonybrook University), Traugott Sheytt (Technische Universität Berlin), Nira Yamachika (Orange County Water District), and Donald Wilkison (USGS) for help in sample collection. Financial support for this research was provided by the National Science Foundation (Grant BES 0303627), the National Science and Engineering Research Council of Canada, and the University of California Toxic Substances Research and Teaching Program.