Bacterial Community Structure and Physiological State within an Industrial Phenol Bioremediation System

Abstract

The structure of bacterial populations in specific compartments of an operational industrial phenol remediation system was assessed to examine bacterial community diversity, distribution, and physiological state with respect to the remediation of phenolic polluted wastewater. Rapid community fingerprinting by PCR-based denaturing gradient gel electrophoresis (DGGE) of 16S rDNA indicated highly structured bacterial communities residing in all nine compartments of the treatment plant and not exclusively within the Vitox biological reactor. Whole-cell targeting by fluorescent in situ hybridization with specific oligonucleotides (directed to the α, β and γ subclasses of the class Proteobacteria [α-, β-, and γ-Proteobacteria, respectively], the Cytophaga-Flavobacterium group, and the Pseudomonas group) tended to mirror gross changes in bacterial community composition when compared with DGGE community fingerprinting. At the whole-cell level, the treatment compartments were numerically dominated by cells assigned to the Cytophaga-Flavobacterium group and to the γ-Proteobacteria. The α subclass Proteobacteria were of low relative abundance throughout the treatment system whilst the β subclass of the Proteobacteria exhibited local dominance in several of the processing compartments. Quantitative image analyses of cellular fluorescence was used as an indicator of physiological state within the populations probed with rDNA. For cells hybridized with EUB338, the mean fluorescence per cell decreased with increasing phenolic concentration, indicating the strong influence of the primary pollutant upon cellular rRNA content. The γ subclass of the Proteobacteria had a ribosome content which correlated positively with total phenolics and thiocyanate. While members of the Cytophaga-Flavobacterium group were numerically dominant in the processing system, their abundance and ribosome content data for individual populations did not correlate with any of the measured chemical parameters. The potential importance of the γ-Proteobacteria and the Cytophaga-Flavobacteria during this bioremediation process was highlighted.

The efficient remediation of xenobiotic pollutants by microbial communities remains a major challenge to microbial ecologists and process engineers alike since bioremediation solutions are based upon the coupling of mechanical engineering with biological diversity and functionality. However, whilst the design and implementation of process engineering solutions are relatively well established, a lack of accurate descriptions of microbial diversity and functionality tends to limit efficient bioremediation. The ability to monitor diversity structuring, stability, and long-term resilience during process management are key requirements for monitoring and predicting bioremediation efficiency. These shortfalls in the understanding of microbial community dynamics and process events are constantly reenforced by reference to the ecological “black box” of microbial remediation systems. However, due to their inherently high biological activity, wastewater processing systems provide an ideal resource for the analyses of microbially based remediation per se and for investigating community structure and succession. The system we have investigated is characterized by relatively high microbial activity, permitting estimation of both community diversity and physiological state in the presence of variable levels of toxicants, the latter driving microbial community structure, adaptation, and succession.

Phenolic wastes are a major class of xenobiotic pollutants from industrial processes such as coking, industrial resin manufacture, and petroleum-based processing (45). Despite the widespread involvement of microbes in the remediation of the diverse array of phenolic compounds produced (35, 36), little is known of the diversity of organisms which fulfill the process, or how the balance between pollutant loading and treatment efficiency is maintained with respect to microbial community dynamics. A historic limitation in our understanding of phenol and other bioremediation systems is that we have only been able to measure gross biological and chemical determinants within the process (15, 41). However, as methodologies improve, the factors that modulate community structure, functionality, and resilience within component microbial consortia can be understood (13, 43).

The deficit in knowledge stems from the fact that many of the microbes which fulfil particular processes may not have been isolated in the laboratory (reviewed in reference 3) or may have specific community associations which prevent the isolation of pure cultures for analysis. Nonetheless, considerable insight to specific remediation processes has been gained by culture isolation and the genetic characterization of important pathways (9, 19, 33, 40). The application of molecular techniques facilitates analyses of environmental samples by the profiling of phylogenetic diversity based on 16S rDNA to assess community structure and succession (14, 21, 28) or the whole-cell targeting of specific taxa or individual organisms by fluorescent in situ hybridization (FISH) (12). These culture-independent molecular studies can provide a more complete understanding of microbial community composition than pure culture studies in that they can demonstrate the in situ presence (and potentially physiological status) of key taxa by ribosome counting (29) and may direct isolation efforts toward these key groups for further laboratory characterization. Ultimately, the aim of a combined molecular characterization of in situ communities and subsequent targeted isolation strategies is the examination of in situ distribution of organisms, their specific physiological function, and their potential for manipulation to optimize community processes.

Typically, detailed analyses of bioremediation communities have focussed on constructed laboratory bioreactors or the major biological reactor area (e.g., 13, 40, 43). In the operational industrial processing system we have studied, several distinct compartments are present. These compartments are of high volume, ca. 1,000 m³, and are linked via a series of pipelines over the extent of the industrial plant. The complex engineering solution to this wastewater treatment involves the collection of coking effluent from a steel manufacturing plant and ground water runoff, mixing of effluent in intermediate holding tanks and inlet control tanks, and finally transfer to a highly aerated Vitox airlift bioreactor prior to discharge (Fig. 1).

FIG. 1.

Schematic diagram of the wastewater-processing system under investigation. P1 A and B, waste-receiving reservoir section 1, subsections A and B; P2 A and B, waste-receiving reservoir section 2, subsections A and B; INTER, intermediate reservoir; B, BITMAC influent; INLET, holding and mixing reservoir for VR; VR, Vitox biological reactor; TIDAL, tidal storage tank prior to discharge. Connecting pipelines and distances are shown in bold whilst the volume of each processing section is indicated in italics. Approximate flow rate into the biological reactor (VR) is also indicated.

To better understand the structure and activity of microbial communities in response to the contaminant loading and its subsequent effect on the biological component, we have applied culture-independent methods as a primary characterization to directly examine the response of the bacterial community in situ to pollutant loads in each of the treatment compartments described. These studies established, at the community level, that the distribution and physiological state of the microbes between compartments correlated with the dominant pollutant present, the phenolic xenobiotics. At a higher resolution, specific bacterial groups had a physiological status which strongly correlated with key aspects of the chemical composition of the effluent (e.g., phenolic and thiocyanate concentration), indicating a potential relationship between the functionality of these defined groups and process chemistry. Molecular analyses demonstrated the presence of highly structured communities which presumably had evolved in response to differential selective pressures within the treatment compartments. Resolving the extent of community diversity and physiological state provides the basis for more-specific analyses and monitoring of process efficacy.

MATERIALS AND METHODS

Sampling site and determination of chemical parameters.

Samples were taken from an undisclosed industrial Vitox wastewater-processing system within the United Kingdom (Fig. 1). Typical operating levels of phenolic species for remediation, as determined by gas chromatography, were between 250 and 500 mg liter⁻¹ in the treatment system and fell to less than 5 mg liter⁻¹ as a result of biological remediation in the tidal discharge holding tank. Data sets, where available, were supplied for the processing compartments under study by the operators and encompassing the chemical determinations of total phenolics, pH, thiocyanate, free cyanide, ammonia, total organic carbon, and nonvolatile matter concentrations.

Sampling and method for nucleic acid extraction.

Large-volume samples (up to 5 liters) were collected from each processing compartment in May 1998 (Fig. 1) and were thoroughly mixed prior to subsampling. Subsamples of 30 ml were then removed, and 20 ml of each suspension was collected onto sterile 0.2-μm-pore-size Durapore filters (Millipore Corp.) and stored at −70°C prior to analysis. Total nucleic acids were extracted directly from filters by the proteinase K-sodium dodecyl sulfate-cetyltrimethylammonium bromide protocol (4), were further purified by phenol-chloroform-isoamyl alcohol (25:24:1) extraction, were precipitated with 0.7% (vol/vol) volumes of ice-cold isopropanol, were washed in 70% (vol/vol) ethanol, were air dried, and were resuspended in 50 μl of Tris-EDTA (10 mM Tris, 1 mM EDTA; pH 7.4).

16S rDNA amplification and DGGE analyses.

For denaturing gradient gel electrophoresis (DGGE) analyses, a 200-bp product spanning the V3 region of the 16S rDNA was amplified from all nucleic acid samples by PCR by using primers targeted to the V3 region of the 16S rDNA essentially as described elsewhere (27) with the modification that the forward primer was targeted to the exact location of the EUB338 probe binding site (primer sequence 5′ ACTCCTACGGGAGGCAGC 3′). Approximately 50 ng of template was amplified with 2.5 U of AmpliTaq (Sigma Chemicals, Dorset, United Kingdom), 1 pM of each primer μl⁻¹, 200 μM of each deoxynucleoside triphosphate, and 1.5 mM Mg²⁺. The PCR protocol was optimized and performed on an MJ Tetrad PCR machine (MJ Research Instruments, Watertown, Mass.) with reaction conditions of 95°C denaturation for 120 s, followed by 35 cycles at 95°C for 60 s, 60°C for 45 s, and 72°C for 90 s, and a single step of 72°C for 30 min.

DGGE analysis was performed by loading ca. 1 μg of PCR-amplified DNA product onto a 10% (wt/vol) acrylamide gel containing a denaturant gradient of 30 to 60% (100% denaturant consisted of 7 M urea and 40% [vol/vol] formamide) parallel to the direction of electrophoresis by using the D-Code system (Bio-Rad, Hercules, Calif.). Gels were electrophoresed at 60°C at a constant voltage of 85 V for 16 h prior to being silver stained by the SILVER SEQUENCE method (Promega Corp., Madison, Wis.). After silver staining, gels were washed in distilled water and were subsequently scanned prior to data analysis (see below). All gels were standardized by the addition of a ladder generated by mixing PCR products of the 16S rDNA gene from a range of cultured isolates from the wastewater system encompassing the minimum and maximum running distances of the range of amplicons to be analyzed. Each standard was loaded onto the outermost lanes of each gel as well as the central lane to normalize the running distance of the amplicons of interest and to correct for gel-to-gel variation and distortion. Gel documentation and band profiling was obtained by scanning the gels at a resolution of 340 dpi by using an Epson flatbed transmission scanner followed by band and profile analyses using Phoretix 1D analysis software according to the manufacturer's instructions (Phoretix International, Newcastle upon Tyne, United Kingdom). Dendrograms for comparisons of DGGE samples were generated by using peak position matching utilizing the Dice coefficient and trees generated by the unweighted pair group method using mathematic averages algorithm programs integral to the commercial profile analysis software.

FISH, microscopy, and image analysis.

The remaining 10-ml volume from the 30-ml subsamples described above were fixed at 4°C for 30 min in ice-cold paraformaldehyde (to a final concentration of 1% [vol/vol] buffered with phosphate-buffered saline [Difco]) and postfixed by the addition of an equal volume of cold ethanol prior to storage at −20°C. For FISH, replicates of 20-μl volumes of fixed cell suspensions were spotted onto clean multiwell slides (ICN Laboratories) and were air dried for 20 min. The probes employed in this study were as follows: EUB338, specific for most organisms in the domain Bacteria; ALF1b, BET42a, and GAM42a, corresponding to the respective subclasses within the Proteobacteria; CF319a, specific for members of the Cytophaga-Flavobacterium cluster; and Ps, specific for most members of the rRNA group 1 pseudomonads, with the exception of Pseudomonas putida, all as reviewed in Amann et al. (3), Daims et al. (10), and Schleifer et al. (30). All hybridization and wash conditions were followed as documented (2, 30), except that the competitor oligonucleotides (2) in the case of the BET42a and GAM42a probes were omitted in order to optimize the accuracy of the rRNA content determination (see below) by maximizing target availability. The possibility that this strategy led to coprobing (through single-mismatch probe sequences) was discounted by statistical analyses of the data which revealed no positive correlation between the two sets of probe counts or the fluorescence determinations (see below). All probes were synthesized with a 5′ Cy3 modification (MWG-Biotech AG, Ebersburg, Germany) and applied separately to each sample. After hybridization, all samples were washed in distilled water and counterstained with 1 μg ml⁻¹ DAPI (4′,6′-diamidino-2-phenylindole; Sigma chemicals) for 20 min to aid cell localization prior to mounting in ProLong antifade medium (Molecular Probes, Ltd., Eugene, Oregon).

Microscopic determinations of cell numbers (DAPI) and oligonucleotide probe-positive cells (Cy3) were performed by using a Nikon Eclipse E600 microscope equipped with a specific DAPI filter set and a G-2A filter set for determination of Cy3-labelled cells. A minimum of 10 fields of view were enumerated for each replicated sample and were counted in duplicate by two independent operators. For calculations of community composition, total DAPI counts were performed for parallel samples where no probe was applied and were used as the total count determination. This approach discounted the possibility of the Cy3 oligonucleotide label absorbing the DAPI counterstain through fluorescence transfer (a phenomenon regularly observed in the samples). Absorbance of the DAPI fluorescence underestimated the total cell count for DAPI and Cy3 dual-stained samples. Therefore, estimates of the percentage of the community assigned to each probe was calculated on the basis of the averaged Cy3 probe counts relative to independent DAPI counts. For quantification of probe fluorescence, random fields of view (accumulating a minimum of 500 probe-positive cells) were digitized by using a 16-bit charge-coupled device camera (Digital Pixel Imaging, Brighton, United Kingdom) with a constant integration time of 2 s. Variation in camera performance was monitored by acquiring and analyzing images of standard 2-μm particles (Becton Dickinson Corp.) on a routine basis to account for signal drift. All captured images were background corrected, segmented, and measured as detailed previously (44). Cell intensity distributions for each probed sample were measured in parallel to RNase-treated and probed samples (100 μg of RNase A ml⁻¹, 37°C for 1 h) to establish background levels which were subtracted from the final fluorescence values obtained for probed cells. In practice, all probe-positive cells had a fluorescence value at least three times that of the RNase-treated and probed background control samples.

Statistical treatment.

In order to examine relationships between community composition and fluorescence determinations with respect to treatment chemistry, all data was log transformed to check for normality and was analyzed for significant correlations within the Minitab package (Minitab 12; Minitab Inc.). Only those relationships are described which were significant at the 5% level (P < 0.05) and which could be obtained over a minimum of five treatment compartments (to account for low abundances in certain treatment compartments or difficulty in applying image analysis algorithms in the case of large cell aggregates).

RESULTS

DGGE 16S rDNA fingerprints of processing compartments.

Molecular analyses by DGGE fingerprinting within the total processing system, of which the Vitox reactor (VR) was the biological remediation point, revealed diverse bacterial communities (Fig. 2). Fingerprinting indicated that the processing sections could be classified based upon bacterial community structure within the system. For example, two distinct community structures were present within the two separate waste collection compartments (P1 and P2). Profile analyses of the P1 and P2 compartments (Fig. 3) indicated that the collection compartments contained communities which were more similar within their respective subcompartments (e.g., P1A and P1B) than between the two separate waste collection compartments. This was determined by the formation of two main clusters at the 35% similarity and by the further classification of the A and B subcompartments within these two respective groupings (Fig. 3). Downstream of these collection compartments, the bacterial community changed markedly and became more uniform (Fig. 2 and 3) as evidenced by the DGGE profile analysis of the intermediate holding tank (INTER) and the bioreactor inlet mixing tank (INLET), which were found to be ca. 95% similar by profile clustering (Fig. 3). Furthermore, the microbial composition of these two compartments was considered to be relatively stable and strongly influenced by selection imposed by the process chemistry since input from a secondary waste source (BITMAC) into the INLET compartment had no influence on the community diversity within the INLET when compared to the INTER compartment (Fig. 2 and 3). In the VR bioreactor communities (Fig. 2), profiles were 60% similar to those of the INTER and INLET tanks, indicating some conservation of the diversity detected within the INTER and INLET tanks but also a certain degree of community composition change (Fig. 3). Habitat-specific profiles were also observed in the discharge tank (TIDAL) that receives the remediated effluent prior to environmental release (Fig. 2 and 3).

FIG. 2.

DGGE profiles obtained for all the treatment sections (corresponding to Fig. 1) after gel digitization and profile extraction. DGGE profile peaks represent band positions and intensities within the gel; the top of the gel is represented by lower pixel positions whilst increasing pixel position values represent distance down the gel.

FIG. 3.

Unweighted pair group method using mathematic averages dendrogram of DGGE profiles obtained in Fig. 2 for each compartment within the processing system, indicating the similarity between the fingerprints by pairwise comparisons of DGGE band presence and position.

Population structure within the processing system by whole-cell hybridization using Proteobacteria group-specific probes.

The EUB338 probe was applied to investigate the relative distribution of the domain Bacteria. However, variance in performance of EUB338 (see Discussion) was observed; therefore, estimates of the true structure of bacterial communities were determined by the analysis of data obtained with probes of greater taxonomic resolution (ALF1b, BET42a, GAM42a, and CF319a) against independent DAPI counts. Total DAPI counts varied between 9.1 × 10⁶ and 2.9 × 10⁷ cells ml⁻¹ within the processing system (Fig. 4a), which were principally accounted for by the summation of counts obtained with the applied group-specific probes (Fig. 4b). However, in the P2 A compartment, significant overestimates occurred in the summation of the percentage of the community with respect to the total cell count obtained (ca. 230% of the total DAPI cells were detected by the specific oligonucleotides). For the Proteobacteria group-specific probes, the majority of the cells were assigned to the β or γ subclass, with less than 4% of the total cells reacting with the α probe (Fig. 4b). DGGE profiles demonstrated that the microbial diversity within, but not between, the P1 and P2 compartments were similar, and group-specific probing reinforced this. For example, the P1 A and B subcompartments contained intermediate levels of γ-Proteobacteria (22 and 18% of total DAPI counts, respectively) and few β-Proteobacteria (undetectable and <1.0% of total DAPI counts, respectively). In contrast, the P2 A and B compartments exhibited large increases the percentages of β- and γ-Proteobacteria (Fig. 4b). For the INTER and INLET compartments, the microbial diversity was similar when assessed by FISH in that each compartment contained low levels of α-Proteobacteria and appreciable levels of β and γ Proteobacteria (i.e., α, 1.7 to 2.2%; β, 17 to 29%; and γ, 30 to 54%) in each of the compartments (Fig. 4b). The VR community was substantially comprised of cells which could be assigned to the γ group (36%) and a lower proportion of β-Proteobacteria (19%). This structure changed slightly when compared to the community resident in the TIDAL compartment where similar levels of β-Proteobacteria could be detected (17%) but the γ-Proteobacteria accounted for only 10% of the total cells present.

FIG. 4.

Total cell counts obtained by DAPI staining and community structure by fluorescent in situ hybridization counts (% of total DAPI count) for populations within processing system compartments. (A) Total DAPI count. (B) Community composition based upon the percentage of the total DAPI cells assigned to the respective subclasses within the Proteobacteria by specific probes α 1b, β 42a, and γ 42a. (C) Group-specific probes: CF319a for members of the Cytophaga-Flavobacterium cluster; Ps, percentage of pseudomonds in the total DAPI cells detected.

FISH analyses with higher-resolution oligonucleotides.

Group-specific oligonucleotides (Fig. 4c) further increased the resolution for assessing the microbial diversity structure within the processing system. Cytophaga-Flavobacteria and pseudomonad probes revealed that the majority of the cells within the processing system were assigned to the Cytophaga-Flavobacterium group, with a maximum of 84% recorded in P2 B (Fig. 4c). Cells hybridizing with the Pseudomonad probe were detected consistently (ca. 20 to 40% of the total cells) in the INTER compartment and further downstream (Fig. 4c), after establishing low populations in most of the wastewater capture areas (i.e., P1 A, P1 B, and P2 B compartments).

Relationships between bacterial distribution, ribosome content, and chemical parameters.

When relating process chemistry data (phenolics, thiocyanate, total organic carbon, ammonia, free cyanide, and nonvolatile matter) with the total DAPI count and community composition in each processing section, only the total cell abundance was found to correlate significantly (P < 0.05) with any of the measure chemical parameters (Fig. 5). Bacterial abundance exhibited a linear inverse relationship with total phenolic concentration over all the treatment compartments, indicating a strong modulation of total cell count by the primary pollutant.

FIG. 5.

The relationship between total DAPI count and phenolic concentration measured throughout the processing system. Treatment compartment values are indicated as follows: P1A, ●; P1B, ■; P2A, ○; P2B, □; BITMAC, ×; INTER, ◊; INLET, ▴; VR, ▵; TIDAL, ⧫.

In combination with the total counts and probing with group-specific rRNA, quantitative image analyses of whole-cell fluorescence was applied to provide tentative estimates of in situ physiological status of each targeted microbial population. Of the chemical parameters measured, only the total phenolic concentration significantly correlated with those cells that probed positively with the EUB338 probe (R² = −0.71, P = 0.032) (Fig. 6a), indicating that for these populations ribosome content was tightly coupled to total phenolic concentration. The γ-Proteobacteria ribosome content positively correlated to total phenolics (R² = 0.937, P = 0.019) and secondly with thiocyanate (R² = 0.915, P = 0.029) (Fig. 6b and c).

FIG. 6.

(A) EUB338 probe fluorescence (as an indicator of ribosome content) for probe-positive members members of the domain Bacteria within the process compartments and its relationship with total phenolic concentration over the nine compartments. (B) Relationship between probe γ 42a fluorescence and phenolic concentration. (C) Relationship between probe γ 42a fluorescence and thiocyanate concentration. Note that the γ 42a data is restricted to five compartments (see Methods and Materials). Only relationships significant at P < 0.05 are indicated. Treatment compartment position in all the plots are indicated as follows: P1A, ●; P1B, ■; P2A, ○; P2B, □; BITMAC, ×; INTER, ◊; INLET, ▴; VR, ▵; TIDAL, ⧫.

DISCUSSION

Bacterial communities within a specialized wastewater processing system were analyzed by 16S rDNA DGGE for total-community fingerprints and group-specific FISH probes in order to assess the community structure and place this within a chemical process framework. In order to circumvent difficulties in interpreting the changing community composition (e.g., when using highly resolved probes and primers) and its association with process chemistry, we selected a generalized approach which used group-level FISH probes and PCR primers for 16S rRNA-V3 region analyses by DGGE. By using this strategy, the DGGE analyses provided data on the presence and extent of sequence diversity and an indication of approximate community structure, whereas the targeting of communities by whole-cell probing allowed analysis of the actual distribution of component groups within the identified communities. Whilst each method differs in the way in which diversity is detected, we observed that gross shifts in community structure profiles (DGGE) were mirrored by changes in whole-cell distributions observed by FISH (i.e., the distinction between P1 and P2 compartments and the similarities between INTER and INLET compartments). Specifically, we assume that this congruence was due to the relatively low and distinct diversity present in these highly specialized communities which are more than likely dominated by only a few distinct groupings of organisms. This was evidenced by the relatively low number of highly defined DGGE bands and the dominance of a low number of distinct probe-positive morphologies (data not shown) present in virtually all the highly polluted compartments upstream of the VR.

A key observation from the FISH studies was the overestimation of probe-delimited organisms in some of the samples when analyzing the cumulative percentage of group-specific rRNA probe-assigned organisms (e.g., up to 230% of DAPI stained cells for the P2 B compartment) but was directly attributed to the presence of large aggregates of cells that could not be dispersed adequately. One further observation was the overestimation of cells present when expressing the group-specific probe counts relative to the number of cells which probed positively with EUB338. Whilst the summation of the probe-positive counts accounted for the majority of DAPI cells within the treatment compartments as a percentage value (Fig. 4b and c), the number of EUB338-positive cells accounted for 40 to 90% of the cells present with a mean probing value of ca. 60% of the total DAPI count. This clearly indicated that the EUB338 count data tended to underestimate the potential number of probe-positive cells within the sample. The β and γ probes have a single mismatch, which could account for higher estimations of each group by coprobing in the absence of unlabelled competitors relative to the EUB338 count. Such interference was discounted, since no statistically significant correlation between the counts obtained by each probe over all the samples was observed. Further, overestimations occurred even when only one of the groups was seemingly detected (e.g. P1 A and P1 B sections) (Fig. 4). Recent evidence suggests that these group-specific probes probably underestimate total members of the group (16). An explanation of the inadequacy of the EUB338 probe to describe all the group-specific probe-positive cells may be the limitation in specificity of the probe to the bacterial domain. Related studies (10) have begun to identify the lack of reactivity of EUB338 with groups such as Planctomyces and Verrucomicrobium. In order to examine potential sources of error within our studies, we examined the EUB338 probe specificity against a general sequence set of 16S rDNAs to establish a potential level of error and then to further classify its homology with full-length sequences for organisms belonging to the α-, β-, γ-Proteobacteria and Cytophaga-Flavobacterium group.

Preliminary analysis indicated that of 27,772 bacterial 16S sequences available, 58% contained the EUB388 consensus sequence, and 72% contained the sequence if a single mismatch was incorporated. Although this is more than likely an underestimate due to partial sequence inclusion and potential sequence errors around the EUB338 target site, the major portion of the sequences do contain the complete V3 region data. At the more resolved group level, full-length sequence alignments indicated that more than 90% of the sequences for organisms contained within the α, β, and γ subclass groupings contained exact matches with the EUB338 target sequence. Occasional mismatches were observed within the available Cytophaga-Flavobacterium sequence that aligned to the CF319a probe (ca. 60% of available sequences indicated an exact EUB338 consensus sequence but many sequencing errors were apparent within this region). Although limitations of the EUB338 probe can be demonstrated and clearly include errors due to inclusion of partial sequences and sequencing errors, increases in database size and quality and the specific application of the probes in complex systems will develop and refine probes and help to resolve these overestimation issues as potential errors in the in situ analyses of complex communities (e.g., 26).

In the highly specialized communities studied in this investigation, DGGE analyses highlighted the relatively low diversity of most treatment compartments as compared to more complex environments such as soil and water. The DGGE microbial diversity community fingerprint for each process compartment using general bacterial primers indicated that the bands detected (ca. 20 to 50) were dominated by less than 20 bands. This diversity was geographically distinct (i.e., the different communities in the wastewater-receiving compartments P1 and P2 or the similar communities in the INTER and INLET), suggesting strong community structuring within individual compartments. We hypothesize that detoxification and remediation of key pollutants probably occur in all compartments owing to the high microbial load, distinct community structuring, and detectable rRNA contents observed. However, because of the compartmentalized community structure in the different process sections, it is unclear whether a functionality (process) is conserved through the changing community structure (compartment) or whether the functionality changes with changing community composition. Recent evidence supports the view that functionally stable bioreactors can be maintained even in the presence of a varied community structure (13). With the combination of approaches described for the microbial system under investigation, we can begin to address this central hypothesis of community structure and functionality at an industrial scale, principally by extending the analyses to more powerful techniques for single-cell gene expression studies (6) in tandem with sequence data for key remediation pathways.

Whole-cell hybridization confirmed that the γ-Proteobacteria were prevalent in most sections of the processing system, whilst the β-Proteobacteria exhibited localized abundance, as in the BITMAC compartment. These data are in agreement with previously published work where β- and γProteobacteria comprise a large fraction of the bacteria in wastewater treatment plants (31, 37, 38). One striking observation was the prevalence of the Cytophaga-Flavobacterium group in the majority of the processing compartments. The Cytophaga-Flavobacterium group is known to be present in wastewater (25) and lake and marine ecosystems (16). Based on their numerical abundance, assessed by culture-independent methods, the Cytophaga-Flavobacterium group appears to play an important process role or occupy a common niche within this industrial wastewater treatment plant. Indeed, analyses of the system by microbiological agar isolation yielded few colonies typical of the Cytophaga-Flavobacterium group organisms (<0.04%), suggesting bias against these organisms during culture surveys with general agars (unpublished data). The exact functional nature and diversity of the members of this group within the processing system awaits resolution by the application of probes of higher resolution (32) and further characterization of their phylogeny and physiology through both culture and culture-independent analyses.

In order to gain some insight into the culture-independent status of organisms within the system, we correlated the chemical process data with bacterial community composition and ribosomal content within the probe-targeted populations. Of the relationships observed, the most significant was that total phenolics, the major pollutant, modulated microbial densities and ribosomal content (relative fluorescence): both decreased significantly (P < 0.05) as phenolics increased. High relative phenolic concentrations were, potentially, the limiting factor to the efficient remediation. However, for the Cytophaga-Flavobacteria (the numerically dominant group), their abundance and physiological status did not correlate with any measured variable (total phenolics, pH, thiocyanate, free cyanide, ammonia, total organic carbon, or nonvolatile matter concentrations). The independence of their distribution and physiological status suggests that members of this group exhibit strong competitive advantages, such as the tolerance of high levels of toxicity of the major pollutants, and/or may not conform to the rRNA-physiological status relationship as found in other organisms (for an example, see reference 22). Although their exact functional nature within this system has yet to be resolved, it has been noted in other systems that some members of this group can degrade phenol-based compounds (24).

The ribosome content of the γ subclass organisms correlated strongly with total phenolics and thiocyanate concentrations through the compartments included in the data analysis, indicating that this subclass could contain the predominant process degraders. However, no correlation was recorded for the pseudomonads (a major constituent of the γ-Proteobacteria). This observation was unexpected, since pseudomonads are well characterized in their ability to degrade environmental phenol contamination, and the genetics of phenol degradation has been described by a variety of strains (1, 5, 7, 8, 18, 23, 34, 39). However, the lack of correlation we observed in situ between phenolics and pseudomonad physiological status may not be unexpected, since culture isolation conditions for bioremediation organisms are highly selective (11, 42) and the catabolic genes involved are often associated with the horizontal gene pool (17, 20, 46). Examining the physiology of distinct and diverse groups at relatively high resolution, as employed for the pseudomonads here, may, for example, overlook the presence of a small fraction of the group which possess, or have acquired, specific functionality in the selective environment. In associated investigations, we have applied nonselective culture isolation from the VR and observed that less than 10% of the isolated pseudomonad strains were able to utilize phenol as a sole carbon source (unpublished data). Clearly, identification of the key functional organisms within a defined treatment process must be based upon both their presence, activity in situ, and physiological traits detected by molecular methods. Of the remaining probe-delimited groups examined, the β-Proteobacteria appeared to have a distribution and activity that correlated to ammonia concentration. However, the direct correlation determined by data analysis was dependent on a single population present in the BITMAC compartment where ammonia concentrations were unusually high (175 mg/liter compared with an average value of 50 mg/liter). This was not investigated further, but analysis of temporal samples would be required to better resolve any relationship attributable for these important organisms in wastewater treatment systems.

The use of combined rapid community fingerprinting, whole-cell hybridization, and image analysis has allowed us to target and identify the key functional components of a microbial biotreatment plant. These analyses, together with the more specific application of probes and culture efforts on a temporal scale, will allow the future deconstruction of the microbial consortium into those members which have key process activity and those which may provide biosensors or indicator strains by which the efficacy of the system can be assessed.