The activated sludge samples were collected from the wastewater treatment plant in Pocatello, Idaho, USA. To construct a microcosm, 1 liter of sludge samples was added to an Erlenmeyer flask and kept at 20˚C. Fresh air was pumped into each microcosm at 27.5 ml ・ second⁻¹ to simulate the aeration process in the sewage treatment plant.

E. coli strains used in this study were EHEC strain EDL 933 provided by the National Food Safety & Toxicology Center at Michigan State University, and non-EHEC strain K-12 received from American Type Culture Collection (ATCC). E. coli cells were cultured in Luria-Bertani (LB) medium and collected at log phase. E. coli cells were kept in 0.1% peptone solution at 4 × 10¹⁰ ml⁻¹ colony forming unit (CFU) [25] . Ten milliliters of E. coli solution was added to each microcosm every 2 days. One set of 3 microcosms included one EHEC added microcosm, one K-12 added microcosm and one control microcosm without any bacteria added. The experiment was repeated in triplicate.

Samples were collected from one set of microcosms at day 0, 2 and 4 for total bacterial counts. The total bacterial counts were performed using the serial dilution method on LB agar plates [26] . Sample collection for T-RFLP analysis was done according to Li et al. [27] . Briefly, 1 ml of sludge sample was collected as the control sample before the sludge was added to each microcosm at day 0. At day 10 and 25, 1 ml of suspended sludge sample was collected from each microcosm. DNA was extracted using the FastDNA^® SPIN Kit for Soil Kit (MP Biomedicals, Santa Ana, CA, USA). Each DNA sample was equally divided into 3 aliquots and stored at −20˚C before PCR amplification.

At day 0, 1 ml of sewage sample was collected from one of the control microcosms for 18S rDNA analysis. At day 53, 1 ml of sewage sample was collected from one of the three EDL 933-added sewage microcosms for 18S rDNA analysis. The FastDNA^® SPIN Kit for Soil Kit (MP Biomedicals, Santa Ana, CA, USA) was used to extract total DNA from activated sludge samples.

PCR amplification of ciliate specific18S rDNA was carried out with the following primers: 384F (YTB GAT GGT AGT GTA TTG GA) and 1147R (GAC GGT ATC TRA TCG TCT TT) [28] . PCR amplification reactions were carried out in 50 µl reactions containing 1 × PCR buffer, dNTPs at 0.8 mM each, 2 mM MgCl₂, each primer at 0.4 μM, 1.25 U of Takara Ex Taq polymerase (Clonetech, Mountain View, CA, USA), and 0.2 μg non-acetylated BSA (Promega, Madison, WI, USA) in a PCT-200 Peltier Thermal Cycler (MJ Research Inc., Watertown, MA, USA). Amplification conditions were 94˚C for 1 min, held at 80˚C for addition of dNTPs (hot start), followed by 30 cycles of 94˚C for 45 seconds, 55˚C for 1 min, and 72˚C for 1.5 min, followed by a final extension at 72˚C for 10 min. Amplified 18S rDNA gene fragments were cloned into pGEM^®-T Easy Vector (Promega, Madison, WI, USA) and transformed into E. coli DH5α according to the manufacturer’s instructions. Recombinant colonies from each source were identified from Ampicillin/X-gal plates. The plasmid DNA was extracted using the alkaline-lysis mini-preparation method [29] . The plasmids were grouped according to restriction fragment length polymorphism patterns following double digestion by RsaI and MspI (Promega, Madison, WI, USA) and visualization by 2% agarose gel electrophoresis [30] . Plasmid inserts from representative clones of each unique RFLP group were sequenced with the SP6 primer (TACGATTTAGGTGACACTATAG). Basic Local Alignment Search Tool (BLAST) was used to identify ciliate species from the 18S rDNA gene sequences based on the highest identity scores. Selected clones with different BLAST search results were used for sequencing with the T7 primer (TAATACGACTCACTATAGGG). Vector NTI Suite 9 (Invitrogen Corporation, Carlsbad, CA, USA) was used to align contiguous sequences. RDP3 was used to detect possible chimeric sequences from clone libraries [31] . Assembled contig sequences were searched against GenBank accessions via BLAST, and the matched ciliate species with highest identity scores were identified. MEGA 5 was used to construct the phylogenetic tree using the maximum likelihood method [32] . The sequences of 18S rDNA clones and the closely related ciliate species sequences obtained from the NCBI database were used in the phylogenetic tree with Paramecium tetraurelia as the outgroup.

T-RFLP analysis was conducted according to Liu et al. [33] by using fluorescently labeled 18S rDNA primers: 5’FAM-384F and 5’HEX-1147R. PCR conditions were the same as the 18S rDNA clone library experiments. PCR products were purified and washed with 200 μl TE in the Montage^® PCR spin columns (Millipore, Bedford, MA, USA) and resuspended in 30 µl TE. Aliquots of 15 μl were digested with MspI and RsaI separately and incubated overnight at 37˚C. The digested mix was amended with 2 μl of 2.5 M sodium acetate, pH 5.2, and precipitated with 60 μl isopropanol overnight at 4˚C. The pellet was recovered after 10 min of centrifugation at 16,100 × g and washed with 70% ethanol, vacuum dried and resuspended in 5 μl TE. T-RFLP analyses of the fluorescently labeled fragments were performed with an ABI 3100 automated capillary DNA-sequencer (Applied Biosystems Instruments, Foster City, CA, USA) at the ISU Molecular Research Core Facility. T-RFLP analysis expedited (T-REX) software [34] was used to analyze the T-RFLP tabulated data exported from Peak Scanner^TM software (Applied Biosystems Instruments, Foster City, CA), using the additive main effects and multiplicative interaction model (AMMI) [35] . The terminal restriction fragments (T-RFs) were processed with the following parameters: Noise filtering (peak area standard deviation multiplier = 1), T-RF alignment (clustering threshold = 0.5), T-RFs omitted if they occurred in less than 2% of samples, “sample name” and “day” were chosen as environments. One-way analysis of variance (ANOVA) was used to compare the numbers of T-RFs among different samples [36] .

Partial 18S rDNA sequences were deposited in Gen Bank under accession numbers JX667685 through JX667704.

The total bacterial cell count from the original activated sludge was 1.7 × 10⁴ ml⁻¹ from one set of 3 microcosms. The total bacterial cell count from the K-12-added microcosm changed from 3.4 × 10⁸ ml⁻¹ (with K-12 cells added), to 1.2 × 10⁶ ml⁻¹ after 2 days, to 1.5 × 10⁴ ml⁻¹ after 4 days. The total bacterial cell count of EDL 933-added microcosm changed from 5.8 × 10⁸ ml⁻¹ to 1.5 × 10⁶ ml⁻¹ after 2 days, to 1.3 × 10⁴ ml⁻¹ after 4 days. This could indicate that the EDL 933 and K-12 cells were removed at approximately the same rate in microcosms. Based on this result, fresh E. coli solution at 4 × 10¹⁰ ml⁻¹ CFU was added to each microcosm to reach 10⁸ ml⁻¹ final concentration every 2 days.

Based on the ciliate18S rDNA sequencing results, several ciliate species were identified (Table 1). Epistylis wenrichi was the most commonly found ciliate in the original sludge sample before the treatment (70% in 79 clones),

^aTotal number of clones that shared the same RFLP patterns with the sequenced clones.

but it was not detected in the EDL 933-treated sludge sample. Prorodon teres was common in both the EDL 933-treated sludge sample (52% in 56 clones) and the untreated original sludge sample (30% in 79 clones). From the EDL 933-treated sludge sample, 41% were Vorticella fusca, 5% were Dexitrichides pangi and 2% were Opisthonecta henneguyi. Those three species may be resistant to EHEC as they remained at very low numbers in untreated sludge sample. Their phylogenetic relationship is demonstrated in Figure 1, along with EHEC resistant ciliates Vorticella microstoma, Platyophyra bromelicola and Colpoda aspera identified by Ravva et al. [13] .

The T-RFLP AMMI analysis of ciliate 18S rDNA genes (Figure 2) showed differences between the E. coli EDL 933 strain, the K-12 strain, and the no bacteria added control microcosms, based on MspI and RsaI restrictions of triplicate samples. The T-RFLP pattern of original sludge sample at day 0 was different from the patterns of no bacteria added controls at day 10 and day 25, while the patterns from day 10 and day 25 controls were quite close to each other. Without adding new bacteria as food source, the original ciliate population could starve and thus the population would change. At day 10 and day 25, EDL 933-treated ciliate populations were different from K-12-treated ciliate populations (Figure 2). At day 25, EDL 933-treated ciliate populations were more similar to no bacteria added controls, than the K-12-added ciliate populations.

The T-RFs average abundance from each T-RFLP file was also analyzed (Table 2). Base on the one-way ANOVA test, the numbers of T-RFs from the EDL 933-treated samples and K-12-treated samples, and the no bacteria added controls were not significant different at day 10 in RsaI restricted samples (F_2,6 = 1.431, p = 0.31). At day 25, the number of T-RFs from the three types of treatments were not significant different from each other in MspI restricted samples (F_2,6 = 0.54, p = 0.609) and RsaI restricted samples (F_2,6 = 0.201, p = 0.823). Al-

^aNumber of T-RFs from K-12-treated samples in day 10 with MspI restriction (n = 2); ^bNumber of T-RFs from original sample in day 0 (n = 1).

though the AMMI analysis showed that EDL 933-treated ciliate populations were different from the K-12-treated ones, the numbers of T-RFs were not significantly different from each other.