Effect of Sludge Age on the Bacterial Diversity of Bench Scale

Mar 9, 2009 - The changes observed with time are consistent with ecological drift. Sludge age has a subtler and slower effect than we anticipated. How...
0 downloads 13 Views 787KB Size
Download PDF
Environ. Sci. Technol. 2009, 43, 2950–2956

Effect of Sludge Age on the Bacterial Diversity of Bench Scale Sequencing Batch Reactors ALPER TUNGA AKARSUBASI,† OZGE EYICE,† IAN MISKIN,‡ IAN M. HEAD,‡ AND T H O M A S P . C U R T I S * ,‡ Molecular Biology and Genetics Department, Istanbul Technical University, 34469 Istanbul, Turkey, and School of Civil Engineering and Geosciences, University of Newcastle upon Tyne, Newcastle NE3 4RP U.K.

Received September 17, 2008. Revised manuscript received January 26, 2009. Accepted February 6, 2009.

Sludge age or mean cell residence time (MCRT) plays a crucial role in design and operation of wastewater treatment plants. The change in performance, for example micropollutant removal, associated with changes in MCRT is often attributed to changes in microbial diversity. We operated four identical laboratory-scale sequencing batch reactors (two test and two control) in parallel for 212 days. Sludge age was decreased gradually (from 10.4 to 2.6 days) in experimental reactors whereas it was kept constant (10.4 days) in control reactors. The reactor performance and biomass changed in a manner consistent with our understanding of the effect of sludge age on a reactors performance: the effluent quality and biomass declined with decreasing MCRT. The composition of the bacterial and ammonia-oxidizing bacterial communities in four reactors was analyzed using denaturing gradient gel electrophoresis (DGGE), and similarities in band patterns were measured using the Dice coefficient. The overall similarity between the communities in reactors run at different sludge ages was indistinguishable from the similarity in communities in reactors run at identical sludge ages. This was true for both the general bacterial communities and putative AOB communities. The number of detectable bands in DGGE profiles was also unaffected by sludge age (p ∼ 0.5 in both cases). Initially, the detectable diversity of activated sludge communities in all four reactors clustered with time, regardless of their designation or sludge age; however, these clusters were only weakly supported by bootstrap analysis. However, after 135 days, a sludge age specific clustering was observed in the bacterial community but not the putative ammonia-oxidizing bacterial community. The mean self-similarity of each reactor decreased, variance increased, and the number of detectable bands in DGGE profiles decreased over time in all reactors. The changes observed with time are consistent with ecological drift. Sludge age has a subtler and slower effect than we anticipated. However, we postulate that sludge age may be more evident in the taxa occurring below the detection limit of DGGE. New sequencing technology may help us address this hypothesis.

* Corresponding author e-mail: [email protected]. † Istanbul Technical University. ‡ University of Newcastle upon Tyne. 2950

9

ENVIRONMENTAL SCIENCE & TECHNOLOGY / VOL. 43, NO. 8, 2009

Introduction Mean cell residence time (MCRT) or sludge age is a key parameter in the control of many biological treatment systems. It represents the average amount of time that a cell will spend within a system. Consequently, the net growth rate of a given organism must equal or exceed the MCRT if it is to be maintained within a reactor. Downing et al. (1) and Lawrence and McCarty (2) demonstrated how the relationship between growth rate and MCRT was the key to nitrification in particular and reactor performance in general. Latterly, sludge age emerged as an important parameter governing the removal of a wide range of micropollutants, including pharmaceuticals and endocrine-disrupting chemicals (EDC) (3-6). Typically, sludge ages of greater than 10 days are required to ensure micropollutant removal (4). It has been hypothesized that reactors with higher sludge ages maintain a more diverse bacterial population and are therefore more able to degrade certain synthetic compounds (7). The mechanisms governing the microbial diversity of biological treatments systems are still very poorly understood (8, 9), not least because microbial diversity is remarkably difficult to determine (10). Diversity is typically measured using rapid methods such as denaturing gradient gel electrophoresis (DGGE) of PCR amplified small fragments of 16S rRNA genes or cloning and sequencing of larger 16S rRNA gene fragments. These methods take, at best, a random sample of the sequences present in the environment. This in effect restricts such methods to the analysis of the most abundant taxa (11). Consequently, differences in diversity are only detectable if either the composition or relative abundance of the most abundant taxa changes. DGGE has further limitations because without sequencing of excised bands, it reveals essentially no taxonomic information. The investigator sees only the appearance or disappearance of a band on a gel; moreover, diversity may be underestimated if two sequences migrate to the same point on a gel and heteroduplex formation can result in the appearance of novel bands in a DGGE profile that do not represent a real organism. Nevertheless, fingerprinting technologies do allow very large numbers of samples to be processed and lengthy surveys to undertaken and, if applied and interpreted in the correct manner, they are a useful tool in microbial ecology. In a very interesting paper, Saikaly et al. (12) evaluated the effect of sludge age on diversity using terminal restriction fragment length polymorphism (TRFLP), a fingerprinting technique which was interpreted using a number of different indices of diversity. They concluded that diversity was affected by sludge age, but in a counterintuitive manner, with a decrease in sludge age increasing diversity. Unfortunately, some of the comparisons in this study were confounded by time. Thus, we cannot preclude the possibility that the changes in community structure observed were caused by temporal variation rather than sludge age. Some temporal variation due to random births and deaths, sometimes called drift, is probably inevitable, though it may not necessarily occur at rates that can be observed in nature (13). We show that sludge age had a weak, but perceptible, effect on the diversity of detectable bacteria after a number of months of operation and no discernible effect on ammonia-oxidizing bacteria in reactors operated with sludge ages of 10.4, 7.8, and 2.6 days. We speculate that the community dynamics associated with change in sludge age are relatively slow and obscured by temporal effects and the low detection limits of standard molecular methods. 10.1021/es8026488 CCC: $40.75

 2009 American Chemical Society

Published on Web 03/09/2009

Materials and Methods Bioreactor Setup and Sampling Regime. Four identical sequencing batch reactors were set up as duplicate pairs in two separate rooms. Reactor vessels were constructed in transparent acrylic and held a final volume of 5 L of mixed liquor. The reactors were fed artificial wastewater with a concentration of 0.16 g/L peptone, 0.11 g/L meat extract, 0.03 g/L urea, 0.007 g/L NaCl, 0.00 4 g/L CaCl2 · 2H2O, 0.002 g/L MgSO4 · 7H2O, and 0.028 g/L K2HPO4 (prepared by diluting a 100 fold stock in distilled water). The wastewater was stored at 4 °C and fed through commercial 5W UV aquarium filters and delivered by peristaltic pumps (Watson-Marlow, Brendal, MA) accurately controlled ((1% volume error) by P4 conductivity-based level controllers (Harker Electronics, UK). Sludge wastage and effluent drainage were performed using peristaltic pumps (Watson-Marlow) via fixed level calibrated outlets. Accuracy of drawn fluid was (2% volume (mixed liquor) and (2% volume (effluent). Reactors were run on an approximately 6 h cycle comprising 5 h and 10 min of aeration (>6 mg L-1 dO2), 5 min of nonaerated stirring during which the mixed liquor was wasted, and 30 min of quiescent settling followed by 10 min when effluent was withdrawn. There were four such cycles per day. The reactors had a 0.5 day hydraulic retention time. The reactors (two test and two control) were inoculated with the same sample of activated sludge from Washington treatment plant (County Durham, UK) treating domestic wastewater and operated in parallel for 212 days. The reactors were run without sludge wastage for 7 days before the sludge age was gradually decreased to 10.4 days over a 7 day period. In the two test reactors (numbered 1 and 3) the sludge age was reduced to 7.8 days on day 65 (over a one week period) and on day 104 the sludge age of these reactors was further reduced to 2.6 days. The two control reactors (numbered 2 and 4) were maintained with a constant sludge age of 10.4 days. Toward the end of the experiment (day 160), allylthiourea was added to reactors 2 and 1 in an attempt to eliminate the ammonia oxidizing bacteria. However, nitrification subsequently inexplicably failed in all reactors at all sludge ages and the experiment was terminated. Samples were taken for chemical, gravimetric, and culture independent analysis of bacterial communities. Analytical Methods. During the operation of the reactors, temperature and pH were monitored daily. Feed and effluent COD were analyzed once a week. Mixed liquor-suspended solids (MLSS) were measured every day. Five day biochemical oxygen demand (BOD5), ammonia, and total Kjeldahl nitrogen (TKN) (because urea was the nitrogen source) (using a Gerhardt reflux and distillation system, Gerhardt Ko¨nigswinter, Germany), and nitrite and nitrate (using a DX-100 Ion Chromatograph; Dionex, CA), were analyzed weekly. The sludge volume index (SVI) was measured for the first 80 days only. All analyses were carried out according to the Standard Methods for the Examination of Water and Wastewater (14). Sampling and DNA Extraction. Duplicate samples of 50 mL were taken from the reactors every second day. Samples were stored at -20 °C. DNA was extracted from 1 mL aliquots of the stored samples using a Fast DNA Spin Soil Extraction Kit (Q-Biogene, UK) following the manufacturer’s instructions. Polymerase Chain Reaction (PCR). Partial 16S rRNA genes of ammonia-oxidizing bacteria (AOB) were amplified from the extracted genomic DNA by PCR using a MJR Triple Head cycler (MJ Research, Waltham, MA). A nested amplification procedure was used for AOB. Primers pA and pH′ were used in the first round of amplification followed by CTO189fGC and CTO654r in the second round (Table 1). For the second round of amplification, 1 µL of product from the first round reaction was used as template. The first round of amplification comprised initial denaturation at 95 °C for 3 min, followed by 35 cycles of 95 °C for 30 s, 40 °C for 30 s, 72 °C for 1 min, with a final elongation step at 72 °C for 5

TABLE 1. Primer Sequences Used in This Studya oligonucleotide primer pA pH′ CTO654r CTO189fa (A+B+C) 3 b 1

primer sequence 5′-3′

annealing temp (°C) ref

AGA GTT TGA TCC TGG CTC AG AAG GAG GTG ATC CAG CCG CA CTAGCYTTGTAGTTTCAAACGC

40 40 57

15 15 16

GAGRAAAGYAGGGGATCG ATTACCGCGGCTGCTGG CCTACGGGAGGCAGCAG

57 57 57

16 17 17

a Numberings list the corresponding positions in the E. coli 16S rRNA (18). b Addition of a high melting temperature GC-clamp: (5′-CGCCCGCCGCGCGCGGCGGGCGGGGCGGGGGCACGGGGGG-3′).

min. Conditions for the second round of amplification were identical except that an annealing temperature of 57 °C was used. Primers 1 and 3 of Muyzer et al. (16), (Table 1) were used for touchdown amplification of the V3 region of bacterial 16S rRNA genes. The following thermocycler program was used for amplification of bacterial 16S rRNA gene fragments, 95 °C for 3 min followed by 24 cycles of, 95 °C for 30 s, 60-55 °C for 30 s reducing 1 °C every other cycle, 72 °C for 30 s and then 15 more cycles at 55 °C annealing; the program was ended with a final elongation step at 72 °C for 5 min. All PCR reactions were performed in a total volume of 50 µL. PCR reactions contained 10 µM of each primer, 0.2 mM dNTPs, 1 U BioTaq enzyme in the buffer provided by the manufacturer (Bioline, London, UK), and 1 µL DNA template. BioTaq enzyme was treated with UV light for 10 min prior to PCR to remove contaminating DNA from the Taq DNA polymerase preparation. PCR products were stored at 4 °C prior to DGGE analysis. Denaturant Gradient Gel Electrophoresis (DGGE). DGGE was performed with an Ingeny DGGE system (Ingeny, Netherlands). PCR products were loaded on 1 mm thick 10% (w/v) polyacrylamide (37.5:1 acrylamide:bisacrylamide) gels containing a 35-60% linear denaturant gradient for AOB PCR products and a 40-65% gradient for bacterial PCR products (100% denaturant is 7 M urea and 40% (v/v) deionized formamide). All gels were run at 60 °C and 100 V for 975 min in 1X TAE buffer (40 mM Tris-acetate, 1 mM Na-EDTA, pH:8.0). Gels were stained with distilled water containing SYBR Gold (Sigma-Aldrich, Inc., St. Louis, MO; 1:1000 diluted), and images were taken using a Fluor S imaging system (Bio-Rad, Hercules, CA). A modified silver stain (19) was used to stain all gels for comparison and longterm preservation. Gels were blot dried and scanned using a flatbed scanner. Analysis of DGGE Patterns. Gel images were analyzed using Bionumerics Software Vers.4.1 (Applied Maths, Belgium). Similarity between DGGE profiles was calculated using Pearson product moment correlation coefficients (densitometric curve-based), and similarities in band patterns were measured as Dice coefficients (unweighted data based on band presence or absence). The Dice-based cluster analysis was bootstrapped using Treecon (http://bioinformatics.psb. ugent.be/software/details/3). Mean similarities and associated standard deviations were calculated from the similarity values of matrices of all pairwise similarities between DGGE profiles. Evenness, a measure of the equitability of the abundance of the observed taxa, was calculated using the Buzas and Gibson’s evenness method (20) using PAST (http:// folk.uio.no/ohammer/past/).

Results Sludge age had an effect on the operational performance of the reactors under investigation. However, although we VOL. 43, NO. 8, 2009 / ENVIRONMENTAL SCIENCE & TECHNOLOGY

9

2951

FIGURE 1. Variation in the mixed liquor suspended solids in four reactors run at a sludge age of infinity for 7 days and then 10.4 days until day 68. Reactors 1 and 3 were operated at 7.2 day until day 105 and then reduced to 2.6 until day 212. The sludge age of reactors 2 and 4 was maintained at 10.4 days. The MLSS values changed with the sludge age as expected.

FIGURE 2. Effluent nitrate concentration over the experimental period in the four reactors run at a sludge age of infinity for seven days and then 10.4 days until day 68. Reactors 1 and 3 were operated at 7.2 day until day 105 and then reduced to 2.6 until day 212. The sludge age of reactors 2 and 4 was maintained at 10.4 days. observed changes in both community composition and the number of bands in DGGE profiles, these changes were not associated with changes in sludge age until after over 100 days operation at different sludge ages. Solids. MLSS changed with sludge age in the expected manner. The initial MLSS value of 1500 mg/L decreased until day 100 (Figure 1). At this point, a gradual decrease in MLSS was observed in the reactors in which SRT was reduced to 2.6 days while the control reactors, with a sludge age of 10.4 days, had a relatively constant MLSS. After day 150, the difference in the MLSS concentrations of the reactors at different sludge ages was clear (Figure 1). The SVI was monitored for the first 80 days, and SVI in the reactors at lower sludge age was slightly higher (mean 68.4 mg) than in the high sludge age reactor (61.61 mg), the difference was statistically significant (nested ANOVA, P < 0.001). There was no statistically significant difference in SVI prior to the change in sludge age. Nitrification in the Reactors. Nitrification in the reactors was assessed in terms of nitrate formation and TKN removal. For the first 20 days of the experiment, the effluent nitrate concentration of the reactors increased to 84.56 ( 7.7 mg/L. It then declined until day 38, reaching a value of 12.5 ( 2.55 mg/L. After day 68, reactors 1 and 3, which were held at a MCRT of 7.2 days, had lower effluent nitrate concentrations than that of the control reactors (Figure 2). On day 105, the MCRT of reactors 1 and 3 was again reduced, this time to 2.6 days, and a further reduction in mean effluent nitrate concentration was observed (19.8 ( 8.5 mg/L n ) 16). This was lower than the mean effluent nitrate concentration in the reactors operated with an MCRT of 10.4 days (79.0 ( 18.8, n ) 16, p ) 0.002). 2952

9

ENVIRONMENTAL SCIENCE & TECHNOLOGY / VOL. 43, NO. 8, 2009

The observed TKN removal was consistent with the observed nitrate concentrations. Between the days 7-68, reactors had similar TKN removal efficiencies of 47 ( 6%. On day 105, the MCRT of reactors 1 and 3 was reduced to 2.6 days, and they had 41 ( 12% TKN removal until day 165 while the other reactors which had constant MCRT of 10.4 days had removal efficiencies of 51 ( 7%. After day 165, the TKN removal of all four reactors was just 19 ( 13%. Effect of MCRT on Bacterial Community Composition. The composition of the bacterial communities as a whole, and putative ammonia-oxidizing bacterial (AOB) communities in the four reactors was analyzed using DGGE. The greatest differences in community composition were expected between the control reactors (2 and 4) operating with an MCRT of 10.4 days and the experimental reactors (1 and 3) operating with an MCRT of 7.4 days and 2.6 days. Bacterial community comparisons were conducted for the initial period of operation (all reactors with an MCRT of 10.4 days) and for the period where the MCRT of reactors 1 and 3 was reduced to 7.4 days and 2.6 days (Figure 4a,b). When all reactors were operated with the MCRT of 10.4 days, similarity within replicate reactors (80.41 ( 1.100, n ) 50) was the same as the similarity between experimental and control reactors (80.40 ( 1.098, n ) 100, p ) 0.991). Between days 105 and 212 when reactors 1 and 3 were operated at a MCRT of 2.6 days, the similarity between replicate reactors (57.898 ( 0.957, n ) 200) was again the same (p ) 0.127) as the similarity between control and test reactors (55.950 ( 0.937, n ) 400) even though the mean similarity values had decreased in all cases. As the time frame was further narrowed, we noted that on day 134 there had been 11.5 sludge ages since the control reactors were set to 10.4 days and since the short sludge age reactors were set to 2.6 days. The detectable bacterial diversity in the samples bracketing this date (days 126 and 140) was compared using ANOVA. The mean Dice similarity between reactors at the same sludge age (0.802 ( 0.913; n ) 4) was not statistically distinguishable (P ) 0.604) from the similarity in reactors at different sludge ages (0.824 ( 0.057; n ) 8). Cluster analysis clearly demonstrated that the bacterial communities in the reactors cluster temporally and latterly with respect to sludge age (Figure 4a) into three distinct groups. These large clusters may be an experimental artifact as they coincide with the three gels used to analyze the data. Nevertheless, it is instructive to look at the clustering within each group. The first group represents samples taken in the first 70 days when all the reactors had a sludge age of 10.4 days. The reactors designation had no effect on clustering in this period (Figure 4a, Day 2-70); however, the time sequence (2, 2, 2, 2, 26, 26, 26, 26, 6, 6, 6, 6, 14,14, 14, 14, 70, 70, 70, 70, 54, 54, 54, 54, 35, 44, 44, 44, 44, 35, 35, 35 days) appeared to be nonrandom (runs test, P < 0.001). The second group coincides with the 90 day period after the sludge age of the two test reactors had been reduced; there was no obvious clustering by sludge age or reactor designation (Figure 4a, Day 79-159) again the time sequence appeared to be nonrandomly related to the day of sampling (runs test, P < 0.001). However in the final temporal group more than three months after the sludge age in the reactors was changed (comprising 5 sludge ages at 7.8 days and a further 24 sludge ages at 2.6 days), the reactors cluster by sludge age and reactor designation (Figure 4a, Day 166-212). In this case the sequence of samples appeared random with respect to time (runs test, P ) 0.301). The clusters were not strongly supported by bootstrapping but were observed consistently with a different (Raup and Crick) similarity metric and supported by a simple visual examination of the gels (Figure 5). The rate of drift and stability can be assessed by comparing a reactor with itself at a fixed time (day 2 was selected in our analysis) or by comparing the community profiles on a given

FIGURE 3. Comparison of the rate of drift and changes in stability in the reactors. Drift was evaluated by comparing the community composition in the reactor with the community profile observed at day 2 for bacterial banding patterns (a) and AOB banding patterns (c). Stability was evaluated by comparing the community composition on a given day in a given reactor with the previous sample from the same reactor (self-similarity). Data from the bacterial banding patterns (b) and AOB banding patterns (d) are presented. date with the previous sampling occasion. Drift decreased asymptotically over time in all reactors (Figure 3a and 3c). The reactors rapidly attained an apparent dynamic equilibrium characterized by usually consistent Dice values of about 0.8 (Figure 3b,d) and correlated level of similarity between sampling intervals (Figure 3b,d). The two very large drops in similarity between sampling intervals may be an experimental artifact, as they coincide with the three gels used to analyze the data. Interestingly, during the last period of the experiment, the mean similarity in bacterial DGGE profiles within each reactor was somewhat lower and the variance higher (73.429 ( 8.477, n ) 84) than when all reactors were operated with an MCRT of 10.4 days (78.471 ( 5.598, n ) 40). This was true even of the control reactors where the MCRT was unchanged. When the same analyses were conducted with DGGE profiles from the putative AOB community, the similarity between reactors was apparently associated with changes over time, rather than operating conditions or reactor designation. Initially (days 7 to 59) the similarity in AOB communities between designated replicate control reactors (85.0 ( 8.85, n ) 50) was no greater than the similarity between reactors designated as control and experimental reactors (85.220 ( 8.940, n ) 100, p ) 0.888). When the MCRT was reduced to 2.6 days, the similarity between replicate reactors (54.318 ( 1.361, n ) 200) was again no greater than the similarity between reactors operated at different sludge ages (55.268 ( 1.390, n ) 400, p ) 0.528). Thus, although the similarity declined with time, this decline was not a function of the sludge age. Again narrowing into the samples bracketing the period when there had been approximately 11.5 sludge ages at the long and short MCRT (between days 126 and 140), we found the mean Dice similarities between reactors at the same sludge age (0.804 ( 0.144; n ) 4) was not

statistically distinguishable (ANOVA: P ) 0.230) from the similarities between reactors at different sludge ages (0.883 ( 0.096; n ) 8). Cluster analysis suggested that the AOB communities clustered temporally rather than by sludge age with three distinct temporal groups (Figure 4b). These clusters may be an experimental artifact, as they coincide with the three gels used to analyze the data. Nevertheless, within each group, cluster analysis grouped the observations in a manner that was apparently nonrandom with respect to time in the first and second group (runs test, P < 0.001). In the last temporal group there was no apparent clustering with respect to sludge age, even after 200 days though the clustering algorithm placed the samples in a sequence that appeared random with respect to time (runs test, P ) 0.601). The same qualitative patterns of apparent drift and dynamic stability were observed in the putative AOB as was seen in the bacterial DGGE patterns. The mean self-similarity values of each individual reactor were again lower and the variance higher toward the end of the run (when two of the reactors were operated with an MCRT of 2.6 days; 68.03 ( 16.14, n ) 40) than at the beginning (all reactors operated with an MCRT of 10.4 days; 84.23 ( 15.25, n ) 84) (Figure 3c,d). Band Number. There was a decrease in the number of detectable bands in DGGE analysis of bacterial 16S rRNA gene fragments in all of the reactors over time from a mean of 29.5 ( 1.3) bands after 2 days to a mean of 23.8 ( 2.8 bands 210 days later. There was a weak (R (2) )30%) but statistically significant (P < 0.001) linear relationship with time. However, there was no statistically detectable difference in the number of bacterial bands at high and low sludge age (nested ANOVA, p ) 0.671). Furthermore, there was no statistically detectable difference in the AOB band numbers at high and low sludge VOL. 43, NO. 8, 2009 / ENVIRONMENTAL SCIENCE & TECHNOLOGY

9

2953

FIGURE 5. DGGE profiles of bacterial 16S rRNA genes from the SBRs on day 70, 140, and 212. The sludge age at the time of sampling is shown. Note the general loss of similarity, even among ostensibly identical reactors, over time.

FIGURE 4. Similarity (Dice coefficient) of the DGGE banding patterns for the bacterial (a) and ammonia-oxidizing bacterial (b) communities for reactors run at different sludge ages. The similarity values were clustered using the coefficient and found to cluster by time even when the reactors were operated at differing sludge ages (day 68 onward). The number represent bootstrap values based on 100 replicates. Values of less than 30 are not shown. age (nested ANOVA, p ) 0.586) and no change in band number over time (8.0 ( 0.8 bands after 2 days and 8.8 ( 2.8 bands after 212 days). Evenness. The evenness of eubacterial DGGE profiles was highly correlated between all reactors irrespective of sludge age (reactor 1 and reactor 2, 3, and 4: 0.810, 0.764, 0.901 respectively, reactor 2 and reactor 3 and 4: 0.633, 0.849, respectively, and reactor 3 versus reactor, 4: 0.845). There was a significant change over time in all reactors (runs test < 0.05), but the change was not related to sludge age. The AOB DGGE profiles were neither correlated nor did they change significantly with time.

Discussion Sludge age did have an effect on the detectable diversity of the reactors. Though the effect was only slow to appear and 2954

9

ENVIRONMENTAL SCIENCE & TECHNOLOGY / VOL. 43, NO. 8, 2009

only affected some of the detectable microorganisms, a sludge age effect could be inferred from the clustering of the last few data points in DGGE profiles generated using the bacterial specific primers. Though these clusters were only weakly supported by bootstrapping, they are consistent with the gels themselves and the fact that the samples no long clustered in a sequence related to time. The clustering with respect to sludge age was subtle and required several months of operation before it was detected. We say subtle because, even where clustering with respect to sludge ages was observed, all the reactors clustered at about 60% similarity, a value only slightly lower than the similarity between identical reactors at long and short sludge ages. There was no detectable difference in band number between reactors at different sludge ages, though the number of bands in all reactors declined with time. We suggest that the impact of sludge age was slow to appear because even this modest effect was only apparent after about day 200; 29 sludge ages and 135 days after the sludge age was reduced in two of the, until then, identical reactors. The sludge age effect was not therefore apparent when the similarities were averaged over time. The very slow rate of acclimatization is intriguing suggesting, at the most, mean net growth rates of 0.23-0.15 d-1 (assuming exponential growth from between 1-4000 cells per reactor and a detection limit of ∼108 cells/mL). The complex succession observed in the reactor could be a complicated function of both the initial numbers of different taxa in the reactor and their subsequent growth rates. Moreover, though a sludge age effect was observed, sludge age did not have detectable effects on all organisms. In the bacterial gels a number of bands were common to all reactors, even after 200 days, and the DGGE profiles generated with primers targeting the AOB did not appear to cluster with sludge age at all. It is possible that the addition of the allylthiourea at very end of the experiment may have obscured a very late change in the effect of sludge age. However, our observation was consistent with an earlier study of the effect of sludge age in a full-scale plant in Sweden (21) in which reactors were compared in

parallel, and it was found that a modest variation (from 15 days to 10 days) in sludge age had no detectable effect on the community composition of AOB. Not all earlier work is in agreement with our findings. In particular Saikaly et al. (12) report statistically significant but counterintuitive (higher diversity at lower sludge ages) differences in evenness with sludge age. In the work of Saikaly, sludge ages of 2 and 8 days were compared after 1, 2, and 3 sludge ages (i.e., after 2, 4, 6 and 8, 16, 24 days respectively). However, because there is a statistically significant decline in the number of DGGE bands in all reactors, comparing a short sludge age reactor after 6 days with a long sludge reactor after 24 days would lead one to confound time and sludge age. It is conceivable that the functional instability in the reactors was so great that it somehow obscured dissimilarities in the reactors. However, we already know that functional similarity alone does not guarantee stability in microbial communities (22). It is interesting to note that sludge age affected biomass and effluent quality before an effect on diversity was observed in this study and thus we assume observable in other studies (3-6). Moreover, a difference in the number of bands, which might be expected if diversity was lower at lower sludge ages, was not observed. We speculate that a change in sludge age has effects we cannot detect using this kind of DNA fingerprinting technology. Organisms occurring below the detection threshold of DGGE (∼1% or ∼107 cells/mL) may still profoundly influence the biology of a biological treatment system and yet appear and disappear, unnoticed by the technology we deployed in this study. Important functional groups such as foaming organisms are known to be present at abundances of less than 107cells/mL (23), and micropollutant degrading species within this group (24) will be undetectable by DGGE using primers targeting the general bacterial population. Unfortunately, the importance of this threshold to studies of microbial diversity is not always recognized (25). Recent technical and mathematical developments may improve our ability to observe the diversity of biological communities in engineered systems (26). Studies using novel sequencing technology may not only resolve the issue of undersampling but also obviate the lack of phylogenetic information implicit in DGGE. By using sequence data we will not only be able to observe change (or otherwise) in the less abundant taxa but be able to infer if this change represents differences at the level of species genus or above. The new sequencing technologies are of course expensive, and those wishing to study the effect of sludge age without recourse to such methods might find that changes in community composition in the less abundant members of the community could possibly be inferred from simple culture-based methods. However, even with the crude technology available to us in this study, it appears that at least some abundant taxa occur at a wide range of sludge ages. It would be foolish to discount the possibility that the effects of sludge age on the performance of wastewater treatment plants is caused, at least in part, by a change in activity in some of these organisms. Time appeared to have a stronger influence than sludge age for much of this study. For example, reactor similarity declined over time in a gently asymptotic manner. Time per se is not a mechanism; however, a decrease in richness and similarity over time may also be attributed to ecological drift (13). If this speculation is correct we would also expect a decrease in evenness, as the loss of rare taxa associated with drift leads to an increase in the proportional abundance of the most abundant taxon. In fact no such effect was detected, which may reflect the inability of DGGE to capture subtle changes in community structure. On the other hand, the dynamics of stochastic systems are slow and affected by scale,

and the rate of change attributable to sludge age in this study was also slow, even in a small reactor. Undoubtedly nichebased explanations can also be advanced. McGuiness et al. (27) have interpreted similar findings to mean that reactors are not “stochastic systems exhibiting chaotic behavior”. It would seem reasonable to assume that both selection and stochastic effects are at work. If this is so, it would be highly desirable to be able to describe and predict these dynamics mathematically to allow engineers to translate observations at model scale into practice at large scale.

Acknowledgments We thank Mr David Rayne for excellent technical assistance, Mr. Christine Jeans for skilled help with the preparation of the artwork, and three anonymous reviewers for their helpful and insightful comments.

Note Added after ASAP Publication This paper was published ASAP on March 9, 2009 with the first author’s name spelled incorrectly; the corrected version published ASAP March 20, 2009.

Literature Cited (1) Downing, A. L. P.; Painter, H. A.; Knowles, G. Nitrification in the activated sludge process. J. Inst. Sewage Purif. 1964, 63, 130– 153. (2) Lawrence, A. W. M.; McCarty, P. L. Unified basis for biological treatment design and operation. J. Sanit. Eng. Div., Am. Soc. Civ. Eng. 1970, 96(SA 3), 757–778. (3) Clara, M.; Kreuzinger, N.; Strenn, B.; Gans, O.; Kroiss, H. The solids retention timesa suitable design parameter to evaluate the capacity of wastewater treatment plants to remove micropollutants. Water Res. 2005, 39, 97–106. (4) Mills, L. J.; Chichester, C. Review of evidence: Are endocrinedisrupting chemicals in the aquatic environment impacting fish populations. Sci. Total Environ. 2005, 343 (1-3), 1–34. (5) Jiang, J. Q.; Yin, Q.; Pearce, P.; Zhou, J. A survey of endocrine disrupting chemicals in sewage and a preliminary treatment trial. Water Sci. Technol. 2005, 52 (8), 1–7. (6) Joss, A.; Andersen, H.; Ternes, T.; Richle, P. R.; Siegrist, H. Removal of Estrogens in Municipal Wastewater Treatment under Aerobic and Anaerobic Conditions: Consequences for Plant Optimization. Environ. Sci. Technol. 2004, 38, 3047–3055. (7) Langford, K. H.; Scrimshaw, M. D.; Birkett, J. W.; Lester, J. N. Degradation of nonylphenolic surfactants in activated sludge batch tests. Water Res. 2005, 39, 870–876. (8) Briones, A.; Raskin, L. Diversity and dynamics of microbial communities in engineered environments and their implications for process stability. Curr. Opin. Biotechnol. 2003, 14 (3), 270– 276. (9) Curtis, T. P.; Sloan, W. T. Prokaryotic diversity and its limits: microbial community structure in nature and implications for microbial ecology. Curr. Opin. Microbiol. 2004, 7, 221–226. (10) Curtis, T. P.; Head, I. M.; Lunn, M.; Woodcock, S.; Schloss, P. D.; Sloan, W. T. What is the extent of prokaryotic diversity. Phil. Trans. R. Soc. B 2006, 361, 2023–2037. (11) Woodcock, S.; Curtis, T. P.; Head, I. M.; Lunn, M.; Sloan, W. T. Taxa-area relationships for microbes: the unsampled and the unseen. Ecol. Lett. 2006, 9, 805–812. (12) Saikaly, P. E.; Stroot, P. G.; Oerther, D. B. Use of 16S rRNA gene terminal restriction fragment analysis to assess the impact of solids retention time on the bacterial diversity of activated sludge. App. Environ. Microbiol. 2005, 71 (10), 5814–5822. (13) Hubbell, S. P. , The unified neutral theory of biodiversity and biogeography; Princeton University Press: Princeton, NJ, 2001. (14) American Public Health Association. WPCF, Standard Methods for the Examination of Water and Wastewater, 19th ed.; American Public Health Association: Washington, D.C., 1997. (15) Edwards, U.; Rogall, T.; Blo¨cker, H.; Emde, M.; Bo¨ttger, E. C. Isolation and direct compete nucleotide determination of entire genes. Characterization of a gene coding for 16S ribosomal RNA. Nucleic Acids Res. 1989, 17, 7843–7853. (16) Kowalchuk, G.; Stephen, L. R.; De Boer, W.; Prosser, J. I.; Embley, T. M.; Woldendorp, J. W. Analysis of β-proteobacteria ammoniaoxidizing bacteria in coastal sand dunes using denaturing gradient gel electrophoresis and sequencing of PCR amplified 16S rDNA fragments. Appl. Environ. Microbiol. 1997, 63, 1489–1497. VOL. 43, NO. 8, 2009 / ENVIRONMENTAL SCIENCE & TECHNOLOGY

9

2955

(17) Muyzer, G.; Dewaal, E. C.; Uitterlinden, A. G. Profiling of complex microbial populations by denaturing gradient gel electrophoresis of polymerase chain reaction-amplified genes coding for 16S rRNA. Appl. Environ. Microbiol. 1993, 59 (3), 695–700. (18) Brosius, J.; Palmer, J. L.; Kennedy, J. P.; Noller, H. F. Complete nucleotide sequence of a 16S ribosomal gene from Escherichia coli. Proc. Natl. Acad. Sci. U.S.A 1978, 75, 4801–4805. (19) Felske, A.; Engelen, B.; Nubel, U.; Backhaus, H. Direct ribosome isolation from soil to extract bacterial rRNA for community analysis. Appl. Environ. Microbiol 1996, 62 (11), 4162–4167. (20) Buzas, M. A.; Gibson, T. G. Species diversity: benthonic Foraminifera in Western North Atlantic. Science 1969, 163, 72–75. (21) Hallin, S.; Lydmark, P.; Kokalj, S.; Hermansson, M.; Sorensson, F.; Jarvis, A.; Lindgren, P. E. Community survey of ammonia-oxidizing bacteria in full-scale activated sludge processes with different solids retention time. J Appl. Microbiol. 2005, 99, 629–640. (22) Fernandez, A. S.; Huang, S. Y.; Seston, S.; Xing, J.; Hickey, R.; Criddle, C.; Tiedje, J. How stable is stable? Function and community composition. Appl. Environ. Microbiol. 1999, 65 (8), 3690–3696. (23) Davenport, R. J.; Curtis, T. P.; Goodfellow, M.; Stainsby, F. M. Quantitative use of fluorescent in situ hybridization to examine

2956

9

ENVIRONMENTAL SCIENCE & TECHNOLOGY / VOL. 43, NO. 8, 2009

(24)

(25)

(26) (27)

relationships between mycolic acid-containing actinomycetes and foaming in activated sludge plants. Appl. Environ. Microbiol. 2000, 66 (3), 1158–1166. Yoshimoto, T. F.; Fujimoto, J.; Watanabe, K.; Mizukoshi, H.; Makino, T.; Kimura, K.; Saino, H.; Sawada, H.; Omura, H. Degradation of estrogens by Rhodococcus zopfii and Rhodococcus equi isolates from activated sludge in wastewater treatment plants. Appl. Environ. Microbiol. 2004, 70, 5283– 5289. Bent, S. J.; Pierson, J. D.; Forney, L. J. Measuring species richness based on microbial community fingerprints: The Emperor has no clothes. Appl. Environ. Microbiol. 2007, 73, 2399. Quince, C.; Curtis, T. P.; Sloan, W. T. The rational exploration of microbial diversity. ISME J. 2008, 2 (10), 997–1006. McGuinness, L. M.; Salganik, M.; Vega, L.; Pickering, K. D.; Kerkhof, L. J. Replicability of Bacterial Communities in Denitrifying Bioreactors as Measured by PCR/T-RFLP Analysis. Environ. Sci. Technol. 2006, 40, 509–515.

ES8026488