Results of toxicological testing of Jefferson Parish pilot plant samples.

Five toxicological tests were performed using concentrated drinking water samples collected at a pilot-scale drinking water treatment plant that had streams treated with different disinfectants (no disinfectant, ozone, chlorine dioxide, monochloramine, or chlorine) before treatment with granular activated carbon (GAC). The toxicological tests used in this study were the Ames Salmonella assay, a subchronic in vivo toxicity assay in mice, the SENCAR mouse skin initiation-promotion assay, a rat liver foci assay, and the lung adenoma assay in strain A mice. These tests were conducted to determine the general toxicity and the mutagenic/carcinogenic potential associated with the use of disinfection and/or GAC in the treatment of drinking water. The stability of the mutagenic activity of the samples tested was determined by repeated analysis using the Ames Salmonella assay. Results indicated that the samples remained mutagenic for the duration of the tests. All the drinking water concentrates (4000 X) prepared by the XAD resin adsorption procedure failed to provide statistically significant indication of carcinogenic activity in the SENCAR mouse, rat liver foci, and the lung adenoma assays. However, concentrates of the chlorine, chlorine dioxide, and monochloramine treated waters gave consistent mutagenic responses in the Ames Salmonella assay. GAC was effective for 6 months in removing both the mutagenicity of chlorine-treated water and the potential of water to become mutagenic when treated with chlorine. In the in vivo, subchronic 30-day toxicity test in mice, some statistically significant differences in organ weights and body weights of animals exposed to different concentrates of some of the samples were observed. However, a consistent pattern of these differences indicating overt toxicity was not detected.


Introduction
Complex mixtures of organic substances found in surface water and goundwater, as well as wastewaters, defy complete characterization. The best estimate of toxicological effects associated with the exposure of the human population to organic chemicals is the result of biological tests on experimental animals and/or organisms exposed at levels that ensure probable positive responses. To facilitate biological testing of organic contaminants in drinking waters, a process of isolation/ concentration is necessary because of the large number of compounds present (both identifiable and nonidentifiable) and their trace occurrence (total organic matter is generally < 10 mg/L).
For the biological testing of concentrated drinking waters treated with different disinfectants and/or granular activated carbon (GAC), a series of toxicological tests were chosen to indicate the general toxicity and *U.S. Environmental  mutagenic/carcinogenic activity of these waters. These tests were: the Ames Salmonella assay (1,2), the in vivo toxicity in mice assay, the SENCAR mouse skin initiation-promotion assay (3,4), the rat liver foci assay (5)(6)(7), and the lung adenoma assay in strain A mice (8).

Sample Collection and Preparation
The concentrated drinking water samples to be used for toxicological testing were collected and concentrated at a pilot-scale drinking water treatment plant in Jefferson Parish, LA. A detailed description of the facility has been previously described (9). The influent stream was split into five separate process streams as indicated in Figure 1. These process streams were untreated, disinfected, using ozone, disinfected with chlorine dioxide, disinfected with monochloramine, and disinfected with chlorine. For this study, the actual sampling sites were 2, 3, 4, 5, 6, 20 + 21, and 36 as shown in  Table 1. The contact time for chlorine dioxide, monochloramine, and chlorine streams was increased from 30 min to 2 hr by use of a modified 50-gal stainless-steel drum. Two methods, reverse osmosis (10,11) and macroreticular resin process (12,13), were used to collect and concentrate organic compounds from drinking water before toxicological testing.
Reverse Osmosis. By use of a FT-30 reverse osmosis membrane and a Nafion cation exchange membrane, 2000 gal of drinking water were concentrated in the field by batch concentration as shown in Figure 2. In this way, 500 gal were reduced to 10 gal of concentrate with concurrent removal of sufficient calcium and magnesium salts to prevent precipitation. This process was repeated until a total of 40 gal of 50 x concentrate was obtained. Figure 3 shows how further concentration was accomplished in the laboratory by repeating a modified dialysis desalting process in conjunction with the usual reverse osmosis process until the salt content was reduced to -0.85% NaCl and a final volume of 5 gal of 400 x concentrate was obtained. The final concentrate was sterile filtered into a sterile, stainless-steel container by use of a series of graded filter elements, of which the last element was a sterile 0.2-,um Poll Ultipor cartridge.
During the reverse osmosis process two continuing chemical adjustments were performed as the samples were being concentrated. These adjustments changed the pH of the water to pH 2 by metering dilute hydrochloric acid into each cycle of sample pumped into the concentrate, and softened the water by use of a Donnan dialysis cation exchange unit (Nafion membrane) with a sodium chloride stripping solution (11). The temperature of the sample and concentrate was maintained at 15°C by using heat exchange coils immersed in chilled water baths.
Samples for total organic carbon (TOC) analysis were collected at various stages in the reverse osmosis process to help monitor the progress of the concentration ----Ten -50 Gallon Samples (I X)  and to determine organic recoveries at different steps in the procedure. The initial TOC of influent water was 3.4 mg/L. The mass recoveries, as indicated in Table 2, during field concentrations (1-50 x ) were generally excellent, and all recoveries were in the 83% to 118% range. Most of the measured membrane (Film Tech FT-30) TOC rejection values were between 92 and 99%. A significant reduction of concentrate TOC mass was lost during laboratory workup of the samples (concentrating from 50 x to 400 x ). Macroreticular Resins XAD-2 and XAD-8. By use of two resin-filled glass columns in series and subsequent elution and concentration, 2000-gal samples of drinking water were reduced to 2 gal (1000 x). The types of resins used were XAD-2 and XAD-8. Each resin was cleaned before use (14). The resins were purified by Soxhlet extraction in three separate organic solvents (methylene chlorine, acetone, and methanol) for 24 hr each. A subsample of the column material was then eluted with diethyl ether and the eluate checked before use for purity by using gas chromatographic (GC) analysis. No detectable impurities were found, and the resins were stored at 4°C in methanol until used. An example of a detailed schematic used in the sample collection at sites 2, 3, 4, 5, 6, and 20 + 21 is shown in Figure 4. Each glass column (14 x 49 cm) was filled with 5 L of cleaned resin. Each column pair was rinsed  with 30 gal of water to flush out the methanol before sample collection. Two thousand gallons of water were processed through the column pairs at a flow rate of 9 to 11 bed volumes/hour (0.4-0.5 gal/min). During the sample collection process, the pH of the influent was monitored and maintained at pH 2 by adding hydrochloric acid. Uniformity of the acid addition was enhanced by an in-line static mixer. Samples for TOC analyses were taken at collection increments of 0, 500, 1000, 1500, and 2000 gal for each process stream collected at points A, B, and C as shown in Figure 4. The sampling schematic of site 36, as shown in Figure 5, differed from the other collected samples in that the chlorine stream was rechlorinated after GAC treatment and had an additional contact time of 120 min before sampling. An example of TOC data from these process streams is shown in Table 3. Approximately 40% of the TOC was retained by the resins. After sample collection, the XAD-8 and XAD-2 columns were separated, and each column was filled with acetone, agitated to completely wet the resins, and then allowed to soak for 15 min. The columns were each eluted with three resin bed volumes of acetone. Each column extract was separately evaporated to 4L with Buchler continuous flow flash rotary evaporator. These acetone samples (approximately 1 gal from each resin column) were stored at -20°C. The aqueous content of the samples was analyzed to be approximately 20%. Immediately before use in toxicity tests, aliquots of each sample were concentrated an additional 5 times by Kuderna-Danish apparatus, whereby 10% Emulphor in water was added in small measured aliquots (to minimize precipitation). Evaporation was continued until all acetone was evaporated and a sample of 2% Emulphor aqueous solution remained, leaving a concentrate equal to 4000 x the original water sample.

Testing of Reverse Osmosis Concentrates
Two toxicological tests were performed on drinking water samples collected at sample points 2, 3, 4, 5, and bResin contaminated with methanol (MeOH) even after prerinse with 50 gal of water. 6 ( Figure 1) and concentrated (400 x) by reverse osmosis process, the Ames Salmonella assay and the in vivo toxicity assay.
Ames Salmonella Assay. The Ames Salmonella assay was used as a screening process to determine the potential carcinogenic activity of the samples. The results of the Ames assay are shown in Table 4. No indications of a mutagenic response for strains TA 98 or TA 100 were apparent with any of the reverse osmosis concentrates when tested either in the presence or the absence of a metabolic activation system ( -/+ S9).
In Vivo Toxicity Assay. The subchronic toxicity of the reverse osmosis samples was evaluated in a 30day study using CD-1 mice, 10 males and 10 females per dose group, in which the aqueous samples were administered as drinking water in concentrations equal to the original concentrate (400 x ) and one-fourth of the original concentrate (100 x ). At the end of the exposure period, a necropsy was performed on each animal to examine it for gross pathological changes. The results of this study are shown in Tables 5 and 6. Taste aversion to the reverse osmosis concentrates was not seen in the study. In fact, significant increases in fluid consumption were detected in four of the female exposure groups. Statistically significant differences in the body weights were found at sacrifice of two male and two female exposure groups, but the differences in body weights were caused by differences in initial body weights, since the rate of growth of the mice in each experimental group was the same throughout the study. Alterations in the ratios of organ weights to body weights are also depicted in Tables 5 and 6. Since these data and necropsy reports did not reveal any overt, subehronic toxicity, histopathological examination of major organs was not performed.

Testing of Macroreticular Resin Concentrates
Drinking waters that were collected and concentrated by the macroreticular resin process were used as samples for toxicological testing. An outline of these tests and sample identification is shown in Table 7. As noted,  aDisinfectants were compared to corresponding nondisinfected groups compared to distilled water control group by orthogonal contrasts following ANOVA. T = statistically significant increase or 4 = statistically significant decrease (p < 0.05). ns = no significant difference. Disinfectants were compared to corresponding nondisinfected groups and nondisinfected groups compared to distilled water control group by orthogonal contrasts following ANOVA. 1 = statistically significant increase or 4 = statistically significant decrease (p < 0.05). ns = no significant difference. Nondisinfected all samples tested were collected on three different dates. Ames Salmonella Assay. Since all the tests for toxicity could not be conducted simultaneously, the sta-bility of the toxic potential of the samples was evaluated using bacterial mutagenicity in the Ames assay as an end point. These data are presented in Figure 6. For those samples which were initially positive in the Ames 300. assay, mutagenicity in both TA 98 and TA 100 strains was consistently detectable throughout the course of the toxicological tests when stored at 4°C in 2% Emulphor. Some differences in the levels of mutagenicity were noted, but these differences were probably attributable to between-experiment variability in the assay. Although data are not shown, no differences in mutagenic activity were found when the mutagenic activity of samples stored for 4 months in acetone at -20°C was compared to that of samples stored in 2% Emulphor at 4°C. In addition to the stability studies, other samples collected on four different occasions, spanning a period of 14 months were also tested for mutagenicity in the Ames test. The results oftests on these samples assayed without S9 are shown in Table 8. When assayed with S9, the results of all samples gave fewer revertants/ liter equivalent dose than those without S9. The concentrates of untreated and ozone-treated water at all sample collection dates were nonmutagenic in the Ames test. Monochloramine-and chlorine-treated water concentrates were consistently mutagenic in both TA 98 and TA 100 strains; chlorine-dioxide-treated water was mutagenic only for strain TA 98. The overall order of mutagenic activity was chlorine > monochloramine > chlorine dioxide. Treatment of the chlorine-disinfected water with GAC was initially effective in removing both the mutagenicity and the mutagen-forming potential of the water. After 6 months use, the GAC only partially removed the mutagen-forming potential but was still effective in removing the mutagenicity of the water. The TA 98 data also indicate that after being used for an extended period (14 months) the GAC may become ineffective in removing the mutagenicity of chlorinated water.
Lung Adenoma Assay. One of the toxicological tests used for determining the tumor-initiating potential of the concentrated drinking water samples is the mice lung adenoma assay. Nine concentrated water samples in 2% Emulphor were tested at two dose levels (4000 x and 2000 x ) in both sexes of 6-week-old strain A mice. The three control groups used were a negative control, a vehicle control (2% sterile Emulphor), and a positive control (10 mg urethane/mouse by gavage, single dose). There were 20 animals per group, for a total of 840 mice in 42 groups. After 2 weeks in quarantine, 0.25 mL of sample was administered by gavage three times a week for 8 weeks with toxicological observations for an additional 16 weeks. After lung perfusion, the adenomas were counted independently by two technicians, and histological confirmation was performed on 10% of the positive specimens. The number of animals within each group that had lung adenomas is presented in Table 9. The positive control animals all had lung adenomas with a Values calculated from the linear portion (initial slope) of the dose-response curves. ns = nonsignificant (less than 2-fold above background). bRefers to the date the samples were collected. a mean of 11.5 adenomas per animal, the vehicle control had 0.1 adenoma per animal, and all other treated groups had a mean of -0.4 adenoma per animal.
The body weight changes and the mortality data within the exposure groups and controls gave further indication that no treatment-related effect was evident during the study.
SENCAR Mouse Initiation-Promotion Assay. A mouse skin initiation-promotion assay of the drinking water concentrates (4000 x) was conducted in SEN-CAR mice. Each test sample of 0.5 mL was administered orally three times per week for 2 weeks. Two weeks after the final initiating dose, 1.0 ,Lg of 12-tetradecanylphorbal-13-acetate (TPA) in 0.2 mL acetone was administered by topical application to the dorsal skin of 1/2 of the experimental animals three times per week for 20 weeks, whereas the remaining animals received acetone only (0.2 mL/mouse) for the same duration. The control groups for this study were a vehicle control (2% Emulphor) and a positive control (urethane at 500 mg/ kg). The data presented in Table 10 represent the skin tumor incidence at 30 experimental weeks.
The treatment with the vehicle control (2% Emulphor) yielded 0.09/9 (Study I), 0.25/25 (Study II), and 0.38/36 (Study III) tumors per animal/animals with tumor. Only the group of animals receiving the chlorine/ GAC exposure had more tumors, at 30 weeks into the study (0.33/33), than did the corresponding control group. The high background activity associated with the oral Emulphor treatment, when averaged for the three studies, is consistent with previous results from this laboratory. The positive control group (urethane at 500 mg/kg) had 0.9 to 1.4 tumors per animal, whereas results from earlier studies in this laboratory with the same oral dose yielded 3.0-3.3 tumors per animal after 30 weeks on study. None of the concentrates of disinfectant-treated water induced tumors on the backs of the mice.
Rat Liver Foci Study. The carcinogenic activity of the organic substances present in the concentrated drinking water samples was also evaluated in the rat liver foci assay. This short-term bioassay determines the ability of the sample to initiate -y-glutamyltranspeptidase (GGT)-positive foci (5-7). Ten rats per group Table 9. Results of lung adenoma assay in strain A mice when tested with two dose levels of concentrated drinking water.

Female
Male  were used for each of the test samples, vehicle control (2% Emulphor), and positive control (50 mg diethylnitrosamine/kg body weight). All rats were hepatectomized on day 0 (% partial hepatectomy), and treated 24 hr later by oral administration of the test and control materials. One week later (day 7), the rats started to receive 500 ppm sodium phenobarbital in their drinking water for a total of 56 days. All animals were sacrificed at day 70. Frozen liver sections were prepared for the histochemical detection and quantitation of GGT foci.
Results of these experiments are shown in Table 11. None of the concentrates of alternative disinfectanttreated water initiated an incidence of GGT foci above that of the vehicle control group. The positive control diethylnitrosamine (DENA) induced a high incidence of GGT foci, which indicated that the test was functioning properly. In Vivo Toxicity. Five of the prepared concentrated samples were used in an in vivo toxicity test. Two dose levels (4000 x and 2000 x) of each concentrated sample were administered to 10 male and 10 female CD-1 mice per dose. Each 0.3 mL dose was administered via gavage three times per week for a total of 4 weeks. Body weights were measured once per week for 4 weeks, and mice were observed for gross physical pathological changes. On day 30, all animals were sacrificed, and a gross necropsy was performed. Summary data from this study are shown in Tables 12 and 13. Table 12 indicates that few changes were detected in the chlorine and monochloramine groups when com- pared to the corresponding nondisinfected groups. However, numerous changes were detected when the toxicity of material passing through "spent" GAC was evaluated before and after rechlorination. These changes between the chlorine + GAC ("spent") and rechlorinated chlorine + GAC ("spent") groups are shown in Table 13. The specific organ weights to body weight ratios were larger in the rechlorinated chlorine

Discussion
The two methods (reverse osmosis and macroreticular resins) used for the collection and concentration of these alternative disinfected drinking water samples have been shown to be reliable in the past when preparing samples for identifying specific constituent organic compounds. However, the results of this study suggest that because of concentration factor limitations (maximum 400 x because of salt interferences), reverse osmosis is not a good method for concentrating drinking water for use in toxicological testing. These results also suggest that the concentration of drinking water samples beyond 4000 x or a better collection/concentration method is warranted if the toxicity and the mutagenic/ carcinogenic activity of trace organic pollutants are to be detected in general animal toxicological experiments. This is particularly true for the rat liver foci assay, lung adenoma assay in strain A mice, and SENCAR mouse initiation-promotion assay, since these methods have previously demonstrated their ability to detect the carcinogenic activity of complex mixtures.
The subchronic in vivo studies with CD-1 mice were designed to detect only the overt toxicities of the disinfected streams. The extracts produced little, if any, obvious toxicological effects. If subtle toxicities had occurred, they would not have been detected by the experimental design. The most significant changes in this study occurred between the chlorine + GAC and rechlorinated chlorine + GAC groups of mice. Apparently, some material passed through the GAC treat- aEach high and low dose group was compared to others by sex by Tukey's multiple comparison test following ANOVA. T = statistically significant increase or 4 = statistically significant decrease (p < 0.05). ns = no significant difference when compared to chlorine group. b Indicates significant differences between chlorine + GAC and chlorine + GAC rechlorinated treatment groups. ment that may have exerted a biological effect during rechlorination. It is not clear why the organ sizes increased in the rechlorinated chlorine + GAC groups. The most significant results of all tests performed were those of the Ames Salmonella assay on the concentrates derived from the macroreticular resin process. This assay demonstrated that the concentrated samples (in 2% Emulphor) remained active with respect to mutagenic activity throughout the 15-month experimental period. This finding indicates that 2% Emulphor may be used as a long-term storage matrix for concentrated aqueous samples. In the Ames Salmonella assay, the chlorine dioxide-, monochloramine-, and chlorinetreated water concentrates were consistently mutagenic. Also, according to the Ames assay, GAC is effective in removing the mutagenicity of chlorinated drinking water and the potential of water to become mutagenic when treated with chlorine. However, after 6 months' use in the chlorine stream, the GAC was only partially effective in removing the mutagen-forming potential but was still effective in removing the mutagenicity of the water.
The results of these short-term animal toxicological studies reveal the difficulties in analyzing actual drinking waters for detectable toxic effects.