INTRODUCTION

Top Abstract Introduction Materials & Methods Results Discussion References

Many branches of industry produce waste gases which contain odorous organic and inorganic components. Apart from the conventional thermal and physicochemical techniques for removal of pollutants from exhaust air, biological waste gas treatment is becoming more and more important. This kind of treatment is advantageous in cases in which the recovery of the components (e.g., absorption in liquids and adsorption in solids) or the utilization of a thermal process (thermal or catalytic combustion) is not economical. Today three different process variations for biological waste gas treatment are used: biofilters, bioscrubbers, and trickle-bed bioreactors. In biofilters and trickle-bed reactors, the pollutant-degrading microorganisms are immobilized on a carrier material, whereas in bioscrubbers the microorganisms are dispersed in the liquid phase. Biofilters and bioscrubbers are preferred in industry, while biofilters are common in compost production and sewage plants (10).

Biological waste gas treatment has a long tradition. Already in 1953, a soil system was employed for the treatment of odorous sewer exhaust gases in Long Beach, Calif. (25), and although up to now a lot of efforts have been funneled into process engineering (14, 17, 18, 24), current knowledge of the involved microorganisms is still very limited. Diversity of the microbial communities in the bioreactors for the exhaust gas purification have mostly been analyzed by culture-dependent methods (9, 12, 28, 31). However, there is a large discrepancy between the total (direct) microscopic cell counts and viable plate counts in many ecosystems and every cultivation medium selects for certain microorganisms. Therefore, cultivation-based studies of bacterial populations may give wrong impressions of the actual community structure in an ecosystem (35). A possible means of avoiding qualitative and quantitative errors in the analysis of microbial community structure in complex ecosystems is the use of fluorescently labeled, rRNA-targeted oligonucleotides (5) for the in situ identification and enumeration of bacteria. This method has already been used successfully in complex microbial communities, such as multispecies biofilms (6, 22, 26), trickling filters (27), and activated sludge (37).

The current study was performed with a laboratory-scale trickle-bed bioreactor degrading a mixture of aromatic compounds (Solvesso100). The starter culture for the inoculation of the bioreactor was an enrichment prepared in a fermentor which was itself started with a wastewater sample from a car painting factory as the inoculum and Solvesso100 as the sole carbon source. The goal of our study was to use for the first time fluorescent in situ hybridization (FISH) to investigate the microbial community structure and dynamics in the fermentor and the bioreactor during start-up. One of the open questions was whether the fermentor enrichment, which is done in suspension, indeed selects for those bacteria that later are immobilized in the bioreactor. In the course of this study, new 16S as well as 23S rRNA-targeted probes for phylogenetic groups within the beta subclass of the class Proteobacteria have been developed and applied in order to obtain a higher taxonomic resolution of the molecular techniques. The molecular data were compared to those obtained by traditional cultivation-dependent techniques.

	MATERIALS AND METHODS

Top Abstract Introduction Materials & Methods Results Discussion References

Model pollutant. Solvesso100 (Exxon Chemical GmbH, Cologne, Germany), a mixture of polyalkylated aromatic compounds, was used as a model of a solvent with limited solubility in water. Table 1 shows the typical composition of Solvesso100. It is produced on a large-scale basis from crude oil by catalytic reforming and is used as a solvent in many types of industrial paints, adhesives, and other products.

Starter culture. A sample of wastewater was collected from a car painting facility (BMW AG, Munich, Germany) to prepare a starter culture for the trickle-bed bioreactor. An aliquot of 100 ml (inoculum I) was used to inoculate a 12-liter laboratory fermentor (model L1523; Bioengineering AG, Wald, Switzerland), filled with mineral medium (17). Solvesso100 was delivered as the sole carbon source continuously with a peristaltic pump. For repeated batch fermentations, the following conditions were used: temperature, 30°C; stirring rate, 500 U/min; gas flow rate, 2.5 liters/min; Solvesso100 feed, 5 ml/h. The fermentor was operated under nonsterile conditions and without pH regulation. After 48 days, 3.5 liters of the exponentially growing, Solvesso100-adapted culture (inoculum II) was transferred to the trickle-bed bioreactor.

Trickle-bed bioreactor. For biological waste gas treatment, a laboratory-scale trickle-bed bioreactor was used. It was packed with hydrophilized polypropylene Ralu rings (Raschig AG, Ludwigshafen, Germany). These Ralu rings were perforated hollow cylinders with a length of 18 mm and a diameter of 18 mm. The specific surface area was 320 m²/m³; the porosity was 0.94. The experimental set-up is presented in Fig. 1. The volume of the column was 10.8 liters, and the height was 0.7 m. The synthetic waste gas was produced by evaporating a controlled amount of liquid into a stream of purified water-saturated air. The air flow rates were adjusted by electronic mass flow controllers (Brooks Instrument B.V., Veenendaal, The Netherlands). The gas entered the column continuously from the top. The recirculating mineral salt solution (17) was distributed continuously at the top by a simple liquid distributor and flowed concurrently with the gas downward through the column. The recirculating fluid served both as absorbent and as nutrient medium. The temperature in the bioreactor was kept constant at 30°C. All experiments were performed under nonsterile conditions. During start-up the reactor was operated with a Solvesso100 inlet concentration of 600 mg/m³ and a gas flow rate of 100 liters/min, resulting in a specific pollutant load of 164 g/m³h.

View larger version (35K): [in this window] [in a new window]   FIG. 1.   Schematic diagram of the experimental set-up of the trickle-bed bioreactor. T, temperature sensor; TC, temperature control; FID, flame ionization detector.

Analytics for Solvesso100. The degradation of the model pollutant Solvesso100 in the bioreactor was analyzed with a gas chromatograph (Fractovap model 4200; Carlo Erba Strumentazione, Milan, Italy) equipped with a flame ionization detector. Gas probes were drawn from the center line of the column at four different positions as indicated in Fig. 1 and measured successively. Automated data acquisition was performed by utilizing a multitasking real-time microcomputer system based on a Motorola 68000 microprocessor. The operating system RTOS-UH/PEARL, developed at the Institut für Regelungstechnik of the University of Hannover, Hannover, Germany, was used.

Sampling. Samples were collected from the wastewater of the car factory, from the fermentor after 6, 29, and 48 days, and from the bioreactor after 127 and 227 days. Ralu rings were also taken from the bioreactor. The biofilm from the Ralu rings taken at days 127 and 227 was scratched off the carrier under sterile conditions and resuspended in 35 ml of 0.9 M NaCl solution. Glass coverslips, which had been placed in the bottom of the bioreactor for microscopic observation of biofilm development, were fixed as described previously (6). For in situ hybridization, the samples were fixed for 3 h with paraformaldehyde as described before (1). The samples were stored in a 1:1 mixture of phosphate-buffered saline (130 mM sodium chloride, 10 mM sodium phosphate buffer, pH 7.2) and 96% ethanol at 20°C. The samples were also fixed by the addition of ethanol to a final concentration of 50%. In situ hybridizations with probes EUB338 and HGC68a were performed on such ethanol-fixed samples, whereas paraformaldehyde-fixed samples were used for probing gram-negative bacteria.

Membrane filtration and staining with DAPI. Total cell counts were determined by membrane filtration and staining with 4',6-diamidino-2-phenylindole (DAPI) as described before (35).

Oligonucleotide probes. All probe sequences, hybridization conditions, and references for this study are given in Table 2. Labeling of amino-linked oligo- nucleotides with carboxytetramethylrhodamine-5-isothiocyanate (Molecular Probes, Eugene, Oreg.) or 5(6)-carboxyfluorescein-N-hydroxysuccinimide-ester (Boehringer GmbH, Mannheim, Germany) and purification of the oligonucleotide-dye conjugates were performed as described before (2). Three new probes specific for phylogenetic groups within the beta subclass of proteobacteria were designed by comparative sequence analysis. In situ hybridization conditions for the new oligonucleotide probes were optimized by gradually increasing the formamide concentration in the formamide buffer as previously described (20).

Dot blot hybridization. The specificities of 23S rRNA-targeted oligonucleotide probes were evaluated by dot blot hybridization of reference nucleic acids extracted from 96 pure cultures of bacteria (representing a diverse collection of taxa) with radioactively labeled probes. The extraction of the nucleic acids and their immobilization on nylon membranes were carried out as described previously (29). The membranes were prehybridized in a solution containing 5× SSC (1× SSC is 0.15 M NaCl plus 0.015 M sodium citrate), 7% sodium dodecyl sulfate, 10× Denhardt solution (0.2% bovine serum albumin, 0.2% polyvinylpyrrolidone, 0.2% Ficoll), and 20 mM NaH₂PO₄ and incubated for 5 h at the hybridization temperature. Hybridization was performed in prehybridization solution containing 5 pmol of the ³²P-labelled probe for 3 to 16 h. The membranes were washed twice with 2× SSC-0.1% SDS at hybridization temperature for 10 min. When the hybridization did not show the expected specificity, the washing procedure was repeated at a higher temperature.

Organisms and culture conditions. The organisms used in this study are listed in Table 3. They were grown as indicated in the respective catalogs of strains.

In situ hybridization and probe-specific cell counts. Fixed samples were immobilized on glass slides by air drying, and in situ hybridizations were performed as described by Snaidr et al. (32). Probes BET42a, GAM42a, and BONE23a were used with competitor oligonucleotides as described earlier (23). Slides were examined with an Axioplan microscope (Zeiss, Oberkochen, Germany) using filter sets 01 (for DAPI staining), 09, and 15. For each probe, more than 5,000 cells stained with the probe EUB338 were enumerated. Color photomicrographs were taken with Kodak Panther 1600X films, whereas black-and-white photomicrographs were done with Tmax400 films. Exposure times were 0.01 to 0.06 s for phase-contrast micrographs and 8 to 30 s for epifluorescence micrographs.

Plate counts and cultivation. Serial dilutions in the range of 10¹ to 10¹⁰ were plated in duplicate on yeast dextrose (YD) agar (tryptose [10 g/liter], yeast extract [2.5 g/liter], dextrose [1 g/liter], NaCl [7 g/liter], agarose [15 g/liter]; pH 7.4), mineral medium (17) with succinate (1 g/liter) as the sole carbon source, and malt extract agar (malt extract [30 g/liter], soy peptone [3 g/liter]; pH 5.6) to determine the total viable counts. The plates were scored after incubation at 30°C for 7 days. Ninety randomly chosen colonies were further cultivated for identification by classical methods and in situ hybridization. Fifty-four isolates were from the biofilm sample, and 36 isolates were from the liquid phase of the trickle-bed bioreactor.

Characterization of isolates. The isolates were subjected to the Gram staining procedure described by Eikelboom and van Buijsen (15) and catalase and oxidase tests. Their morphology and mobility were determined by phase-contrast microscopy. For in situ hybridization, the isolates were cultured in YD agar and cells growing in the logarithmic phase at an optical density at 600 nm (OD₆₀₀) of 0.5 to 0.8 were harvested, washed, and fixed as previously described (36). The oligonucleotide probes listed in Table 2 were used for identification of the isolates.

	RESULTS

Top Abstract Introduction Materials & Methods Results Discussion References

Enrichment for Solvesso100-degrading bacteria in a fermentor. The pH of the fermentor inoculum was 7.2, and the dry weight was 265 mg/liter. The total cell count was 6.9 × 10⁷ cells/ml, whereas on YD agar only 1.9 × 10⁴ CFU/ml could be obtained. Solvesso100 was delivered as the sole carbon source continuously to the mineral salt solution in the fermentor. Bacterial growth in the fermentor was observed after an adaption phase of 16 h. The color of the fermentor culture changed during repeated batch fermentation from colorless to yellow, pink, brown, and finally dirty beige. The generation time of the fermentor culture as determined spectrophotometrically during exponential growth was 7 to 10 h. The maximum optical density was approximately 7 without and 12 with added (NH₄)₂SO₄.

Start-up of trickle-bed bioreactor. The trickle-bed bioreactor was inoculated with 3.5 liters of an exponentially growing fermentor culture (4.2 × 10¹² cells) adapted to Solvesso100 for 48 days. The bacteria were allowed to immobilize on the packing material of the bioreactor by recirculating the liquid phase (liquid circulation rate, 300 liters/h) through the trickling filter system. During the first 15 min after inoculation, a rapid decrease in the optical density of the circulation fluid, from an OD₆₀₀ of 3.5 to 2.9, could be observed and then the OD₆₀₀ stabilized at about 2.5 for the next hours. In parallel, dry mass and total cell counts of the circulation fluid decreased by 24%. Five hours after the inoculation of the bioreactor, the OD₆₀₀ started to rise again and reached 4.7 after 2 days before it decreased to low values again (0.5). This oscillation continued for the remainder of the experiment.

Trickle-bed bioreactor performance. The trickle-bed bioreactor was monitored over a total period of 6 months. During this time, operation conditions, e.g., the gas inlet concentration and the gas flow rates, were changed several times in order to investigate the stability of the biological system and to evaluate the effects of various process parameters. The liquid circulating rate (300 liters/h) and the temperature (30°C) were, however, kept constant throughout the experiment. Already a few days after the start-up of the trickle-bed bioreactor, efficient degradation of Solvesso100 was achieved. The measured specific elimination capacity (defined as the difference of gas inlet and outlet concentrations divided by the mean gas residence time) is plotted versus the specific pollutant load (defined as the gas inlet concentration divided by the mean gas residence time) in Fig. 2. Figure 3 shows the degree of conversion, defined as the difference of gas inlet and outlet concentrations divided by the gas inlet concentration versus the specific pollutant load. A maximum specific elimination rate of approximately 80 g/m³h was achieved. Depending on the pollutant load, about 20 to 60% of Solvesso100 was degraded. After a specific pollutant load of about 400 g/m³h, the elimination rate did not increase with an increase of the pollutant load. The system maintained its full degradation capacity during variations in pH and discontinuity in the Solvesso100 supply. The pressure drop remained constant at about 20 to 30 Pa during the observed period.

View larger version (29K): [in this window] [in a new window]   FIG. 2.   Measured specific elimination capacity (defined as the difference of gas inlet and outlet concentrations divided by the mean gas residence time) versus specific pollutant load (defined as the gas inlet concentration divided by the mean gas residence time). The days when the samples for the microbial examinations were taken (days 127 and 227) are indicated.

View larger version (32K): [in this window] [in a new window]   FIG. 3.   Measured degree of conversion (defined as the difference of gas inlet and outlet concentrations divided by the gas inlet concentration) versus specific pollutant load. The days when the samples for the microbial examinations were taken (days 127 and 227) are indicated.

Bacterial community dynamics in the fermentor and the trickle-bed bioreactor as determined by FISH. In the original wastewater sample that was used to inoculate the fermentor (inoculum I), 76% of the microbial cells visualized by DAPI staining were detected by the universal bacterial probe EUB338. Cells detected by probe ALF1b (16%) and probe BET42a (11%) were abundant, whereas only a few cells (<0.1%) were identified as belonging to the gamma subclass of Proteobacteria (Fig. 4). In the samples taken from the fermentor after 6, 29, and 48 days, 68, 77, and 90% of the cells, respectively, could be detected with probe EUB338. The hybridization signals became increasingly brighter. Moreover, the fraction of cells hybridizing with the bacterial probe EUB338 which could be identified with the group-specific probes ALF1b, BET42a, and GAM42a increased considerably. In the samples taken after day 6, a significant increase of bacteria belonging to the gamma subclass of Proteobacteria could be observed. Cells hybridizing with probe BET42a (30%) and GAM42a (31%) were more frequent than cells hybridizing with probe ALF1b (6%). At days 29 and 48, the fermentor was dominated by bacteria showing a positive hybridization signal with probe GAM42a (72 and 80%, respectively). The amount of cells detected with probe ALF1b decreased to <0.1%. Simultaneous hybridization of the sample taken on day 48 with probes Ppu56a, specific for Pseudomonas putida and Pseudomonas mendocina, and GAM42a showed that nearly 100% of the cells hybridizing with probe GAM42a also hybridized with probe Ppu56a (Fig. 5A).

View larger version (34K): [in this window] [in a new window]   FIG. 4.   Comparison of bacterial composition of the samples from the wastewater of a car factory (inoculum I), the fermentor, and the trickle-bed bioreactor (biofilm and liquid phase) as determined by in situ hybridization. Error bars, standard deviations.

View larger version (108K): [in this window] [in a new window]   FIG. 5.   In situ hybridization with fluorescein- and tetramethylrhodamine-labeled probes. Phase-contrast (upper panels) and epifluorescence micrographs (middle and lower panels) are shown for identical microscopic fields. Bar, 5 µm (all photomicrographs). (A) A sample from the fermentor at day 48 (inoculum II) was hybridized with fluorescein-labeled probe GAM42a (middle panel) and tetramethylrhodamine-labeled probe Ppu56a (lower panel). (B) A sample of the trickle-bed bioreactor biofilm at day 127 was hybridized with tetramethylrhodamine-labeled probe BET42a (middle panel) and fluorescein-labeled probe Bcv13b (lower panel).

Design and optimization of oligonucleotide probes SUBU1237, Bcv13b, and ALBO34a. Based on the PROBE DESIGN tool of the software package ARB (33), three probes were designed and optimized to increase the taxonomic resolution of FISH. Probe SUBU1237 was targeted to a site on the 16S rRNA typical for the genera Burkholderia and Sutterella within the beta subclass of Proteobacteria. The probe sequence together with the exact target position is given in Table 2. A recent database check (October 1997) showed that all nontarget sequences had at least one mismatch with probe SUBU1237. The difference in alignment in Fig. 6 displays the locations of mismatches in the 16S rRNA sequences of phylogenetically closely related bacteria. Binding of probe SUBU1237 to these bacteria could be completely prevented by addition of at least 60% formamide to the hybridization buffer (Table 3). Probe SUBU1237 overlaps partially with the already existing probe PPC (11), which was originally designed for Burkholderia cepacia. The modified version is specific for the genera Burkholderia and Sutterella.

View larger version (35K): [in this window] [in a new window]   FIG. 6.   16S rRNA sequence alignment showing target regions of probe SUBU1237 for a selection of reference strains. Nucleotides are only identified for mispairings; pairings are indicated by dots. Lowercase letters indicate weakly destabilizing mispairing. Uppercase letters indicate strongly destabilizing mispairings.

In situ hybridization of the sample taken on day 127 with probes SUBU1237, ALBO34a, and Bcv13b. The sample taken on day 127 was dominated by members of the beta subclass of Proteobacteria. To further specify this group, biofilm and the liquid phase were hybridized with the three newly designed probes. With probe SUBU1237, specific for Burkholderia spp. and Sutterella spp., 92% of the cells belonging to the beta subclass of Proteobacteria showed a positive hybridization signal (Fig. 5B). Simultaneous hybridization with differently labeled probes SUBU1237 and Bcv13b demonstrated that the majority (83%) of the cells hybridizing with probe SUBU1237 also hybridized with probe Bcv13b, which is specific for Burkholderia cepacia and Burkholderia vietnamiensis. Probe ALBO34a hybridized to 3% of the bacteria belonging to the beta subclass. Together probes SUBU1237 and ALBO34a detected 96% of the beta subclass proteobacteria. The results in the liquid phase were similar: nearly all cells (98%) belonging to the beta subclass of Proteobacteria hybridized with probe SUBU1237 (91%) or probe ALBO34a (7%). With probe Bcv13b, 58% of the cells that hybridized with probe SUBU could be identified.

Cultivation and identification of isolates. Samples from day 127 were also plated on YD agar succinate agar, and malt extract agar. Plating efficiencies of the biofilm samples were surprisingly high: 91% on succinate-agar (3.2 × 10⁸ CFU/ml compared to 3.5 × 10⁸ cells/ml as visualized by DAPI) and 68% on YD agar. The plating efficiency in the liquid phase was lower: 36% on YD agar and 15% on succinate agar. Eighty-six randomly chosen colonies (51 isolates from the biofilm sample, 35 isolates from the liquid phase) from the three different cultivation media were identified by classical methods and by whole-cell hybridizations. All isolates, except for one coccus, were rods varying in size and width. They were all oxidase and catalase positive. Gram staining carried out with the isolates showed that 88% (45 out of 51) of the biofilm isolates and 63% (22 out of 35) of the isolates from the liquid phase were gram negative. The identification of the isolates retrieved from the biofilm by whole-cell hybridization showed low complexity. All gram-negative isolates hybridized with probe BET42a, and 5 of the 6 gram- positive isolates hybridized with probe HGC69a. Hybridization of the colonies obtained from the liquid phase showed differences: still, most (20 of 22) of the gram-negative colonies could be detected with probe BET42a, but more colonies, 11 out of 35 (31%), were identified as members of the gram-positive bacteria with high DNA G+C content. One colony hybridized with probes ALF1b, and another hybridized with GAM42a. All 65 isolates from the biofilm and the liquid phase belonging to the beta subclass of Proteobacteria also hybridized with probe SUBU1237, and 64 hybridized with probe Bcv13b.

	DISCUSSION

Top Abstract Introduction Materials & Methods Results Discussion References

The key finding of this study was that in our experiment the fermentor and the trickle-bed bioreactor selected for different bacteria capable of degrading aromatics. Whereas batch enrichment in the fermentor starting from a complex wastewater sample resulted in an almost pure culture of gamma subclass proteobacteria closely related to P. putida or P. mendocina, the community structure of the trickle-bed bioreactor inoculated with the fermentor enrichment returned to high diversity, with dominance of members of the beta and alpha subclass Proteobacteriaa situation, at least on the group level, similar to the one in the wastewater sample used to inoculate the fermentor. That the gamma subclass proteobacteria dominant in the fermentor were rapidly lost in the trickle-bed bioreactor was quite unexpected, since the bacteria in the starter culture were already adapted to the pollutant and immediately started the immobilization when they were transferred to the reactor, as shown by in situ hybridization. Furthermore, it is well known that P. putida and related bacteria readily form biofilms. Banks and Bryers (8) were able to show that a P. putida culture deposited on a developed biofilm of Hyphomicrobium sp. outgrew the hyphomicrobia and became dominant. Recently, P. putida was also visualized as a representative member of the toluene-degrading population in a biofilter for waste gas treatment by scanning confocal laser microscopy using a 16S rRNA-targeted probe (21). In the present study, the dynamics of the bacterial communities were analyzed by FISH and the plate count technique. Even though the results are somehow different, which is not surprising considering the different methods (35), molecular and classical data show the same clear-cut shift from Pseudomonas sp. in the fermentor to a mixture of mainly beta and alpha proteobacteria in the trickle-bed bioreactor.

Obviously, even though the start-up of the trickle-bed bioreactor was eventually successful, the pollutant adaptation in the fermentor did not enrich for high numbers of bacteria with an ability for long-term colonization in this particular trickle-bed bioreactor. This is an important result, since the inoculation of bioreactors with adapted fermentor enrichments is a common practice (14, 16, 19, 23). The growth conditions within the biofilm and the liquid phase of the trickle-bed bioreactor are significantly different from those of the fermentor. A fermentor might therefore in certain cases support growth of fast-growing, planktonic cells, whereas a trickle-bed bioreactor would select for a multispecies microbial community characterized by good immobilization and long-term stability.

The three new oligonucleotide probes for the beta subclass of Proteobacteria supplement a set of 16S rRNA-targeted probes for genera within the beta1 group of Proteobacteria (4). The combination of the nested probes BET42a, SUBU1237, and Bcv13b allowed a rapid and highly reliable identification and enumeration of Burkholderia cepacia and Burkholderia vietnamiensis in the samples of the trickle-bed bioreactor. This application is an example of the successful combination of 16S and 23S rRNA-targeted probes for the nested identification of bacteria in a complex ecosystem (3) and enabled us to identify the dominant groups of the beta subclass proteobacteria. Among the representatives of the beta subclass of Proteobacteria, complexity was quite low. Apparently, the conditions in the trickle-bed bioreactor were so demanding that only a very limited number of bacteria are able to survive.

As already shown in earlier studies (34, 35), the results obtained after cultivation were different from the direct in situ counts. Every cultivation medium is selective; even still, in this case the main Solvesso100 degraders seem to be readily cultivable. The dominance of bacteria belonging to the beta subclass of Proteobacteria and the abundance of significant numbers of bacteria belonging to the gamma subclass of Proteobacteria still are reflected by the colonies presented on the plates. The dominance of bacteria detected by probes SUBU1237 and Bcv13b among the isolates is significantly higher than the in situ counts, while no isolate could be detected with the probe ALBO. One more obvious discrepancy between the two methods is shown in the results obtained with probe ALF. Even though three different media were chosen for the cultivation, no members of the alpha subgroup could be isolated from the biofilm. However, alpha subclass Proteobacteria from the liquid phase could be cultivated in numbers similar to those detected by in situ hybridization.

In this study, it was demonstrated that pollutant adaptation in fermentors might not enrich for those bacteria that establish in trickle-bed bioreactors. Future studies with the molecular and classical techniques used here should examine whether other inocula are more suitable for start-up and how the methods of pollutant adaptation can be modified to select for the right bacteria.