Changes in Microbial Biofilm Communities during Colonization of Sewer Systems

Abstract

The coexistence of sulfate-reducing bacteria (SRB) and methanogenic archaea (MA) in anaerobic biofilms developed in sewer inner pipe surfaces favors the accumulation of sulfide (H₂S) and methane (CH₄) as metabolic end products, causing severe impacts on sewerage systems. In this study, we investigated the time course of H₂S and CH₄ production and emission rates during different stages of biofilm development in relation to changes in the composition of microbial biofilm communities. The study was carried out in a laboratory sewer pilot plant that mimics a full-scale anaerobic rising sewer using a combination of process data and molecular techniques (e.g., quantitative PCR [qPCR], denaturing gradient gel electrophoresis [DGGE], and 16S rRNA gene pyrotag sequencing). After 2 weeks of biofilm growth, H₂S emission was notably high (290.7 ± 72.3 mg S-H₂S liter⁻¹ day⁻¹), whereas emissions of CH₄ remained low (17.9 ± 15.9 mg COD-CH₄ liter⁻¹ day⁻¹). This contrasting trend coincided with a stable SRB community and an archaeal community composed solely of methanogens derived from the human gut (i.e., Methanobrevibacter and Methanosphaera). In turn, CH₄ emissions increased after 1 year of biofilm growth (327.6 ± 16.6 mg COD-CH₄ liter⁻¹ day⁻¹), coinciding with the replacement of methanogenic colonizers by species more adapted to sewer conditions (i.e., Methanosaeta spp.). Our study provides data that confirm the capacity of our laboratory experimental system to mimic the functioning of full-scale sewers both microbiologically and operationally in terms of sulfide and methane production, gaining insight into the complex dynamics of key microbial groups during biofilm development.

INTRODUCTION

Wastewater collection systems, or sewers, consist of an underground network of physical structures-installations composed of pipelines, pumping stations, manholes, and channels that convey wastewaters from their source to the discharge point, usually a wastewater treatment plant (WWTP). Sewer systems thus prevent the direct contact of urban populations to fecal material and potential microbial pathogens, greatly reducing the spread of infectious diseases. Sewers have traditionally been considered only hydraulic transport systems for sewage, although they are in fact “reactors” where complex physicochemical and microbial processes take place. Wastewater microorganisms are diverse and abundant, and they are exposed to a wide range of both inorganic and organic substrates as well as changing conditions along their transport through sewers (1). In this regard, wastewater transport through the pipes facilitates the formation of microbial biofilms that grow attached to the inner surface of sewer pipes (2). Different factors, such as large surface area, low flow velocity near pipe walls, and nutrient availability, may favor microbial colonization of pipe surfaces and biofilm growth. Formation of fully functional biofilms occurs in different steps, from surface conditioning, adhesion of microbial “colonizers,” initial growth, and glycocalyx formation, followed by secondary colonization and growth (3).

Anaerobic conditions in sewer pipes favor the accumulation of both sulfide (H₂S) and methane (CH₄) as end products of different microbial metabolisms, i.e., anaerobic respiration of organic matter by sulfate-reducing bacteria (SRB) and methanogenic archaea (MA), respectively. Both compounds have detrimental effects on the sewer system, with different consequences for both the installation and its surroundings (2). Accumulation of H₂S in the sewer atmosphere causes malodor in the whole system, health hazards due to the well-known toxicity of H₂S, and corrosion of both the inner surface of pipes and the inlet zones of WWTPs (4, 5). H₂S accumulation also impacts the structural integrity of the sewerage by microbial-mediated corrosion processes, which severely affect the performance and cost of downstream processes at the WWTPs (2, 6). Remediation or replacement of corroded pipes requires a high economic investment for large systems, ranging from several hundreds to several thousands of Euros per meter depending on pipe diameter and location depth (7). On the other hand, buildup of CH₄ in sewers results from the activity of MA that colonize inner pipe surfaces and develop within the biofilm matrix under strict anaerobic conditions (8–10). In addition to being explosive at low concentrations, CH₄ is a major greenhouse gas with a life span of ∼12 years and a global warming potential roughly 21 to 23 times higher than that of carbon dioxide (11). Recent reports suggest that CH₄ emissions from sewers contribute significantly to the total greenhouse gas footprint of wastewater systems (12, 13). Accordingly, different mitigation strategies have been used to reduce H₂S and CH₄ production in sewers (14–24).

Although competition between SRB and MA has been reported in some environments such as freshwaters (25), sediments (25), and WWTPs (26), CH₄ production in sewers containing high sulfate concentrations was first detected by Guisasola and coworkers (8). Assuming that SRB and MA may compete for the same substrates (e.g., complex organic matter, acetate, and hydrogen), their cooccurrence in sewer systems is probably the rule rather than the exception, especially considering the large amount of organic matter in wastewater and the prevalence of anaerobic conditions in many sections of sewer networks. In biofilms, this coexistence may be explained by processes of mass transfer of required substrates (e.g., sulfate and organic matter) into the biofilm matrix, which results in a physicochemical stratification along its thickness. Very recently, Sun and coworkers (10) investigated the stratification pattern of SRB and MA in sewer biofilms thicker than 800 μm, locating the former closer to the biofilm surface and locating the latter in greater abundance at deeper, highly anaerobic layers.

Despite these findings, little information is available on the colonization dynamics and activity of SRB and MA relating to biofilm development in sewer systems. Particularly, processes behind early biofilm colonization by SRB and MA in sewer pipes are still not fully understood. In this regard, a better understanding of how these processes take place and how they affect H₂S and CH₄ production rates during biofilm development is necessary to design effective biofilm control strategies for the commissioning of sewers. This information could be crucial to the development and application of optimal control methods to reduce odor, corrosion, and global warming issues generated by sewer biofilms.

The aim of this study was to investigate the initial stages of microbial biofilm development in sewer systems, with a special focus on the interactions between SRB and MA. Biological activities and phylogenetic community structure during the colonization phase were investigated by using a combination of molecular techniques (denaturing gradient gel electrophoresis [DGGE], quantitative PCR [qPCR], and massive parallel sequencing of 16S rRNA genes from target groups) and process data (H₂S and CH₄ production). The work was carried out by using a laboratory sewer pilot plant fed with wastewater that reproduced a full-scale anaerobic pressured sewer. The microbial community composition was compared with that of a biofilm from a full-scale sewer to validate the data obtained from our laboratory experiments.

MATERIALS AND METHODS

Anaerobic sewer biofilm reactor system.

The study was carried out in a specially designed pilot system validated previously, the SCORe-CT method (Sewer Corrosion and Odour Research—Chemical Testing) (27), that mimics the H₂S and CH₄ production capacity of full-scale rising main sewers by reproducing its main characteristics, including (i) hydraulic features, such as hydraulic retention times (HRTs), turbulence, and area-to-volume ratios, and (ii) wastewater characteristics associated with real sewage. The laboratory system consisted of 3 airtight reactors (reactor 1 [R1], R2, and R3), each of them mimicking a section of an anaerobic sewer pipe (see Fig. S1 in the supplemental material). Each reactor had a volume of 0.75 liters, an 80-mm diameter, and a height of 149 mm. The system was fed with fresh sewage (domestic fresh sewage collected in the upstream sections of the sewer network in the municipality of Girona, Spain, close to its source in households) by a peristaltic pump (Masterflex model 7518-10). Sewage was collected on a weekly basis and kept at 4°C to minimize variation in its composition. Wastewater contained 26.5 ± 2.6 mg S-SO₄²⁻ liter⁻¹ and 0.1 ± 0.1 mg COD-CH₄ liter⁻¹. Volatile fatty acid (VFA) and soluble and total chemical oxygen demand (COD) concentrations were 42.3 ± 8.3 mg COD-VFA liter⁻¹, 325.8 ± 40.8 mg liter⁻¹, and 672 ± 93.2 mg liter⁻¹, respectively. Sewage was heated to 20°C before entering the reactors. Magnetic stirrers (Mr Hei-MixS; Heidolph) were used to ensure homogeneous conditions and to produce a shear within the reactors. Wastewater was pumped 15 times a day in uneven periods (between 1 and 3 h). During these intervals, wastewater was transferred from the storage tank to R1 and then from R1 to R2 and finally from R2 to R3 in order to simulate the HRT pattern observed in a full-scale rising main used as a reference sewer pipe, the Radin collector (lat 42.101843, long 3.131631 [L'Escala municipality, Spain]). The Radin anaerobic pipe is 2,930 m long and has a 0.5-m diameter with an HRT of between 3 and 7 h.

Plastic carriers (Anox Kaldnes, Norway) with a 1-cm diameter were clustered on three stainless steel rods inside each reactor to increase biofilm growth surface area and to provide easily extractable biofilm samples. Taking into consideration the reactor wall and carriers, the total biofilm growth area in each reactor was 0.05 m² (area/volume ratio of 65 m² m⁻³). The system was operated continuously for 48 weeks. The colonization period was monitored during the first 12 weeks after start-up of the system. In addition, characterization of mature biofilms was undertaken during the 12th month after start-up. The microbial community composition of mature biofilms was compared to the composition of the biofilm extracted from the upstream reference section of the Radin sewer pipe. A biofilm sample from the full-scale sewer pipe was obtained from a sewer air scour valve that was constantly in contact with the flowing wastewater. The valve was disassembled, and the biofilm grown on its surface was scraped by using a sterile spatula and collected into a sterile Falcon tube containing 5 ml of phosphate-buffered saline (PBS) (137 mM NaCl, 2.7 mM KCl, 10 mM Na₂HPO₄, 1.8 mM KH₂PO₄), in which the collected biomass was resuspended. The sample tube was maintained at 4°C in a portable icebox until arrival at the laboratory (1 h after collection), where it was immediately frozen at −20°C until DNA extraction.

H₂S generation, CH₄ production, and VFA production/consumption in the laboratory system were monitored as the wastewater was transported through the system. Liquid-phase sampling from R3 and offline chemical analyses were done weekly during normal-functioning (NF) tests for the determination of sulfur species (sulfate, sulfide, sulfite, and thiosulfate), CH₄, COD, and VFAs. Sampling hours covered the entire HRT range (3 h to 7 h). Also, 10 batch tests (BTs) were performed to monitor H₂S and CH₄ production by biofilms. Batch tests were carried out once every 1 to 2 weeks. During BTs, the continuous operation of the reactors was stopped. The feed pump was activated for 10 min to ensure that each reactor was filled with fresh sewage. After this, the feed was stopped, and liquid samples were withdrawn every hour for a 3-h period by using a 10-ml syringe connected through a sampling port fitted with a valve and Tygon tubing. Samples were analyzed for sulfur species, CH₄, VFAs, and COD, as described below. Using linear regression, H₂S and CH₄ production rates were calculated from the sampling-point data. A special 6-h batch test was run in order to investigate changes in methane production depending on the presence of sulfate in R1 and R3. Samples were analyzed every hour over a 3-h period for sulfur species and every hour for a 6-h period for methane in order to determine changes in methane production when sulfate was totally reduced to sulfide.

Daily H₂S and CH₄ emissions (calculated from NF test data) were also determined after 1 year of biofilm development to detect changes in activity between early and mature stages of biofilm development in the system.

Chemical analysis.

Dissolved sulfide in R1 and R3 was measured continuously by using an s::can spectro::lyser UV-visible (UV-Vis) spectrometer probe (Messtechnik GmbH, Austria) (28). For the analysis of dissolved sulfur species, 1.5 ml of wastewater was filtered through disposable Millipore filter units (0.22-μm pore size) and added to 0.5 ml preserving solution antioxidant buffer (SAOB) (29). Samples were analyzed within 24 h in an ion chromatograph (IC) with a UV and conductivity detector (ICS-5000; Dionex). VFAs were measured by gas chromatography (Thermo Fisher Scientific) (coupled with a flame ionization detector). For CH₄ samples, 5 ml of sewage was filtered through disposable Millipore filter units (0.45-μm pore size) and injected into vacuumed glass tubes with the help of a hypodermic needle attached to a plastic syringe. After reaching liquid-gas equilibrium inside the tubes, the samples were analyzed by gas chromatography (Thermo Fisher Scientific) (coupled with an flame ionization detector). COD analyses were performed by using a standard photometric test kit with a commercially available reagent (LCK 114; Hach Lang). Absorbance readings were obtained by using an LCK 314 cuvette test with a DR2800 Hach Lang spectrometer. During start-up, Anox Kaldnes plastic carriers were regularly withdrawn to quantify changes in biomass content as a result of microbial biofilm formation. The biomass attached to each carrier was suspended in MilliQ water by vortexing (Genius-3; IKA) until complete detachment occurred (≈2 min). Concentrations of total suspended solids (TSS) and volatile suspended solids (VSS) were analyzed by using standard methods (30). Biomass content was referred to the carrier surface by using values for volatile suspended solids.

DNA extraction.

DNA was extracted from biofilm biomass collected in R1 and from sewage at different-week intervals during the study period. The biomass attached to each carrier was suspended in 5 ml 1× PBS by vortexing (Genius-3; IKA). Suspended biomass from carriers and samples of wastewater (45 ml) were centrifuged at 11,000 rpm for 5 min at 25°C in an Eppendorf 5804R centrifuge equipped with an F-34-6-38 rotor (Eppendorf, Hamburg, Germany). DNA was then extracted from pelleted biomass by using the FastDNA Spin kit for soil (MP Biomedicals, Santa Ana, CA, USA), according to the manufacturer's instructions. Genomic DNA concentrations of biofilm samples were measured by using a NanoDrop 2000 spectrophotometer (Thermo Scientific, Wilmington, DE, USA).

PCR amplification and 16S rRNA gene fingerprinting.

The microbial composition of biofilms formed on carrier surfaces was studied by combining specific amplification of 16S rRNA gene fragments and fingerprinting by denaturing gradient gel electrophoresis (DGGE) (31). Bacterial and archaeal 16S rRNA gene fragments were amplified from DNA extracts by using primer pairs 357F-GC/907R (32) and 109(T)F/515R-GC (33), respectively. PCR amplification mixtures (final volume of 50 μl) contained 10 μl of MgCl₂ buffer (15 mM), 1 μl of deoxynucleoside triphosphates (dNTPs) (10 mM), 2 μl bovine serum albumin (BSA), 1 μl of each primer (10 μM), 0.25 μl of Taq polymerase, and 2 μl of the DNA sample. DNA extracts were diluted with sterile MilliQ water to a final concentration of 10 to 50 ng μl⁻¹ to avoid inhibition of amplification reactions. Sequences of the different primer pairs used during the study and PCR conditions are summarized in Tables S1 and S2 in the supplemental material, respectively.

DGGE analyses were performed with an Ingeny phorU-2 DGGE system (Ingeny International BV, Netherlands). Samples were loaded onto 6% polyacrylamide gels and run with 1× Tris-acetate-EDTA (TAE) buffer using 30 to 70% (bacterial 16S rRNA) and 30 to 50% (archaeal 16S rRNA) linear denaturing gradients of urea-formamide (100% denaturant agent contained 7 M urea and 40% deionized formamide). A molecular ladder composed by a mixture of known small-subunit (SSU) rRNA gene fragments was loaded into all gels to allow intergel comparisons of band migration. Electrophoreses were performed overnight at 60°C at a constant voltage of 120 V. After electrophoresis, gels were stained for 30 min with 1× SYBR gold nucleic acid stain (Molecular Probes Inc.) in 1× TAE buffer, rinsed, and visualized under UV radiation. DGGE fingerprints were analyzed by using GelCompar II (Applied Maths, Belgium). For sample comparison, a presence-absence matrix was used to calculate similarities between patterns, and statistical analysis based on hierarchical cluster analysis was performed with the Dice distance and the unweighted pair group method using average linkages (UPGMA) grouping algorithm.

DNA from excised bands of wastewater samples was eluted as previously described (34). DNA was then amplified by using the same primer pairs (without a GC clamp) and PCR conditions as those described above but sizing down the number of PCR cycles up to 20. PCR products were directly sent to Genoscreen (Lille, France) for sequencing on both strands. Sequences were checked for chimeras by using Uchime (35), aligned by using BioEdit (36), manually curated, and then compared for the closest relatives in the NCBI sequence database (http://www.ncbi.nlm.nih.gov/blast/) using the Basic Local Alignment Search Tool (BLAST) (37).

Real-time quantitative PCR.

Real-time quantitative PCR (qPCR) assays were used to quantify gene copies of bacterial and archaeal 16S rRNA and dsrA functional genes. All qPCRs were run with a Stratagene MX3005P instrument (Agilent Technologies). For all tests, qPCR standards contained a known number of target 16S rRNA genes. qPCR mixtures for bacterial genes contained 15 μl Brilliant III Ultra Fast SYBR green qPCR master mix (Agilent Technologies), 400 nM (each) forward (1048F) and reverse (1194R) primers (38), and 1 μl of template and were adjusted to a final volume of 30 μl with molecular biology-grade sterile water. DNA sample stocks were diluted with water to a final concentration of 10 to 20 ng μl⁻¹. qPCR for archaeal 16S rRNA genes was carried out under the same conditions as those for bacteria but using forward primer 806F (39) and reverse primer 915R (40) and reducing the number of cycles to 35. Quantification of SRB was based on the dissimilatory sulfate reductase subunit A gene (dsrA) according to methods described previously by Ben-Dov et al. (41). Primer sequences, reaction temperatures, R² values, and amplification efficiencies for each qPCR are compiled in Tables S1 and S3 in the supplemental material. All qPCR analyses were carried out according to MIQE rules for quantitative PCR analyses (42), and all essential information is included in this section.

Pyrosequencing and phylogenetic analyses of microbial diversity.

DNA extracts from biofilms at early stages (weeks 1, 5, and 13), mature biofilms (1 year old), and full-scale sewers were analyzed by tag-encoded FLX-Titanium amplicon pyrosequencing at the Research and Testing Laboratory (RTL; Lubbock, TX, USA). Briefly, genomic DNA from biofilm samples was used as a template in PCRs using universal bacterial (28F/519R) (33) and archaeal (341F/958R) (43, 44) primer combinations complemented with 454 adapters and sample-specific barcodes. Raw sequence data sets were preprocessed at RTL facilities to reduce noise and sequencing artifacts, as previously described (45). Demultiplexing according to sample barcodes, sequence quality assessments, chimera detection, and downstream phylogenetic analyses were conducted with mothur (46). Bacterial and archaeal curated sequence data sets were then aligned in mothur by using the bacterial and archaeal SILVA reference alignments, respectively, available at the mothur website (http://www.mothur.org/). Taxonomic classification of bacterial sequences was carried out by using the RDP taxonomy reference database with a cutoff value of 80% for valid assignments. Classification of archaeal sequences was carried out by using the SILVA reference database and taxonomy files using the same cutoff as that used for bacteria (80%). Operational taxonomic units (OTUs) (97% cutoff) and representative sequences of each OTU were delineated and taxonomically assigned by using mothur. For community analysis, the number of sequences in each sample was normalized by using a randomly selected subset of 1,500 sequences (for bacteria) and 6,000 sequences (for archaea) from each sample to standardize the sequencing effort across samples and minimize any bias due to a different number of total sequences. These normalized sequence data sets were then used in mothur to calculate α-diversity indicators of richness (Chao1) and diversity (Shannon) and to calculate community similarity among sites (β-diversity) based on the weighted UniFrac distance (47). Nonmetric multidimensional scaling (nMDS) analysis was performed on the UniFrac similarity matrices to visualize patterns of community composition. The relative abundance of the most populated OTUs (OTUs with relative abundances of ≥4% of total sequences in at least one sample) across samples was visualized as bubble plots by using bubble.pl (http://www.cmde.science.ubc.ca/hallam/bubble.php).

After taxonomic classification of bacteria, sequences affiliated with the class Deltaproteobacteria were selected and further grouped into 149 OTUs (97% cutoff). Representative sequences of each deltaproteobacterial OTU were delineated and assigned by using mothur and then compared for the closest cultured relative by using BLAST. Phylogenetic trees were constructed in MEGA 5 (48) by using representative sequences of abundant OTUs, defined as those having a relative abundance of ≥4% of total deltaproteobacterial and archaeal sequences in at least one sample and the closest cultured representative sequences.

Statistical analyses.

Statistical analyses were carried out by using SPSS software (version 15.0; SPSS, Chicago, IL, USA). Normality of data was assessed by the Kolmogorov-Smirnov test for values obtained for batch testing and inlet wastewater (sulfate and sulfur balance). The correlation between the sulfate concentration in wastewater and sulfate reduction rates was assessed by the Pearson test.

Accession numbers.

Bacterial and archaeal 16S rRNA gene sequences obtained by DGGE fingerprinting were deposited in GenBank under accession numbers KR080151 to KR080166. Pyrosequencing data from this study have been deposited in the NCBI database via the BioSample submission portal (http://www.ncbi.nlm.nih.gov/biosample/) under accession number PRJNA279227.

RESULTS

Differences in sulfide and methane production/emission between young and mature biofilms in laboratory and full-scale sewer systems.

Changes in microbial biomass were continuously monitored for 12 weeks after the beginning of the experiment to assess biofilm formation within bioreactors (Fig. 1A). Initial biofilm growth was detected after stabilization of the biomass content in the range between 2.1 and 3.5 mg VSS cm⁻².

FIG 1.

(A) Temporal changes of microbial biomass in R1, R2, and R3. (B) Sulfide production rates determined in the batch tests on R1, R2, and R3 and sulfate concentrations in inlet wastewater (IW).

The daily profile of H₂S measured by using the s::can spectro::lyser UV-Vis spectrometer probe showed a gradual increase of H₂S production during the first 12 weeks of biofilm development in R1 and R3 (see Fig. S2 in the supplemental material). The higher H₂S production rate determined for R1 than for R3 was probably related to the low sulfate concentration in the wastewater arriving at R3. H₂S and CH₄ production rates were calculated for the same time period to assess the activity of recently formed biofilms. Figure 1B shows the H₂S production capacity within reactors in batch test experiments. H₂S production increased immediately after the start-up of the system. After the second week of operation, the capacity of the biofilm to produce H₂S stabilized at rates of between 3.5 and 7.7 mg S-H₂S liter⁻¹ h⁻¹. Sulfate reduction rates were between 3.2 and 7.7 mg S-SO₄²⁻ liter⁻¹ h⁻¹, which were positively related to H₂S production rates in each reactor (see Fig. S3 in the supplemental material). Differences in H₂S production showed a good correlation with the sulfate concentration in inlet wastewater (Pearson correlation index [R] = 0.881; P = 0.02). Interestingly, from week 8 to week 12, H₂S production in R1 was higher than that in R2 and R3. Regarding CH₄ production, low rates were detected in all reactors during these early stages of development (0.08 ± 0.11, 0.12 ± 0.16, and 0.16 ± 0.16 mg COD-CH₄ liter⁻¹ h⁻¹ in R1, R2, and R3, respectively).

Sulfide emission was measured weekly for 24 h to evaluate the impact of SRB activity in the system, as an accurate representation of full-scale sewer conditions. After the second week of operation, the H₂S emission rate ranged between 195.7 and 388.8 mg S-H₂S liter⁻¹ day⁻¹ (see Fig. S4A in the supplemental material), representing 78.6% ± 14.0% of the inlet sulfate. Therefore, some SO₄²⁻ was still present in the effluent wastewater (75.3 ± 33.0 mg S-SO₄²⁻ liter⁻¹ day⁻¹) because not all sulfate in the influent wastewater was reduced within the system. On the other hand, CH₄ emissions were very low (between 0 and 8.7 mg COD-CH₄ liter⁻¹ day⁻¹) for the first 6 weeks (see Fig. S4B in the supplemental material) but increased to values as high as 44.5 mg COD-CH₄ liter⁻¹ day⁻¹ from week 8 to week 12.

A 6-h batch test experiment was carried out during week 14 (see Fig. S5 in the supplemental material) to assess if CH₄ production was limited by the presence of sulfate. For the first 4 h, the CH₄ production rate in R3 was twice that of R1 (0.37 and 0.88 mg COD-CH₄ liter⁻¹ h⁻¹ for R1 and R3, respectively). Remarkably, CH₄ production increased after 4 h of testing (1.06 and 2.07 mg COD-CH₄ liter⁻¹ h⁻¹ for R1 and R3, respectively), coinciding with the reduction of all sulfate available.

A high level of variability of VFA production rates was observed due to the simultaneous production and consumption of these compounds during batch test experiments (see Fig. S6A in the supplemental material). Nevertheless, VFA production rates were remarkably low for the first 2 weeks of biofilm development. Furthermore, the concentration of VFA exiting the system was higher than those measured in inlet wastewater (see Fig. S6B in the supplemental material).

Comparison of H₂S and CH₄ emissions measured after 1 year of biofilm development with those calculated during the first 3 months of operation in the laboratory suggested similar activities of SRB but clear differences in methanogenesis. After 1 year of growth, emissions of H₂S by laboratory biofilms were slightly different (204.7 ± 14.6 mg S-H₂S liter⁻¹ day⁻¹) from those measured at the initial stage (316.5 ± 61.0 mg S-H₂S liter⁻¹ day⁻¹). This discrepancy may have been caused by differences in sulfate concentrations in the inlet wastewater between the two periods (26.7 ± 2.5 mg S liter⁻¹ and 16.0 ± 1.0 mg S liter⁻¹ during the first weeks and after 1 year, respectively). Regardless of these differences in absolute values, mature biofilms performed better when these concentrations were compared in relative terms (∼80% and 100% of SO₄²⁻ reduced to H₂S during the initial weeks and after 1 year of operation, respectively). In turn, CH₄ emissions largely increased after 1 year of biofilm growth (from 17.9 ± 15.9 mg COD-CH₄ liter⁻¹ day⁻¹ to 327.6 ± 16.6 mg COD-CH₄ liter⁻¹ day⁻¹).

To determine if the high levels of production of H₂S and CH₄ in mature biofilms under laboratory conditions were similar to the emissions of these compounds under natural conditions (e.g., full-scale sewers), we calculated the daily production of both compounds in both systems. Whereas full-scale sewers discharged 4.56 g S-H₂S day⁻¹ m⁻², laboratory systems produced 1.58 g S-H₂S day⁻¹ m⁻². Similar values were obtained for CH₄ production; whereas the full-scale sewer produced 4.24 g COD-CH₄ day⁻¹ m⁻², laboratory systems emitted 1.65 g COD-CH₄ day⁻¹ m⁻².

Changes in the composition of microbial communities during biofilm development.

DGGE fingerprints showed compositional differences between the bacterial community in the inlet wastewater and that of biofilms grown in R1 over the study period (Fig. 2A). Even though several bands were consistently detected at different time intervals, the variation in the banding pattern suggested changes in the composition of bacterial communities during biofilm development. Hierarchical clustering of samples according to the Dice similarity index clearly segregated wastewater samples from laboratory biofilms. Moreover, biofilm samples clustered according to date of collection (e.g., developmental stage). Less variation between wastewater and biofilm samples was observed for archaeal communities, although a similar clustering of biofilm samples according to date was distinguished (Fig. 2B). A total of 16 of the 23 excised bands (9 and 7 bands from the bacterial and archaeal wastewater communities, respectively) (see Fig. S7 in the supplemental material) yielded good-quality sequences. Differences in the bacterial closest relatives identified and band patterns showed high variability of wastewater bacterial communities. On the other hand, the closest relatives of the identified archaea were less diverse, belonging to Methanobrevibacter smithii and Methanosphaera stadtmanae (see Table S4 in the supplemental material).

FIG 2.

Negative images of DGGE gels of 16S rRNA gene fingerprints for Bacteria (A) and Archaea (B) from wastewater and biofilms grown in R1. Hierarchical clustering of samples based on Dice similarity indexes of the banding patterns are also shown. White arrows indicate biofilm samples used for further pyrosequencing analyses (weeks 1, 5, and 13).

Variations in bacterial and archaeal abundance in R1 biofilms during the study period were assessed by qPCR devoted to monitoring biofilm development. Although bacterial 16S rRNA gene copy numbers were always higher than archaeal 16S rRNA copy numbers, both genes showed similar trends in increases of copy numbers for the first 2 weeks of growth, followed by a steady state, which suggested a balanced composition of biofilm communities for the rest of the study period (see Fig. S8 in the supplemental material). Remarkably, dsrA gene abundance showed a time course similar to that of bacterial 16S rRNA genes (see Fig. S8 in the supplemental material), suggesting similar growth dynamics of SRB for the first 2 weeks of experiment.

The compositions of microbial communities from R1 and full-scale sewer biofilms were assessed by massively parallel sequencing to determine whether or not H₂S and CH₄ production rates measured over time were related to compositional changes of bacterial and archaeal biofilm communities. Bacterial and archaeal 16S rRNA gene libraries were constructed by using pyrotags from different samples collected during the study period (week 1, week 5, week 13, 1-year, and full-scale sewer samples). The relative contributions of bacterial phyla changed during biofilm development (Fig. 3A). Furthermore, the composition of the bacterial community in 1-year-old biofilms was clearly different from that of the full-scale sewer system (Fig. 3A). Sequences affiliated with the bacterial classes Bacilli, Fusobacteria, and Gammaproteobacteria progressively decreased during biofilm maturation. It is noteworthy that no sequences affiliated with these classes were identified in the bacterial community from the full-scale sewer biofilm. In turn, sequences affiliated with the class Betaproteobacteria were prevalent in the full-scale sewer biofilm and in R1 samples collected at the first stages of biofilm development (20 to 26% of total sequences), but they showed less representativeness after 1 year of operation (4.9% of total sequences). On the other hand, the prevalence of sequences affiliated with the classes Synergistia and Deltaproteobacteria increased during biofilm colonization, reaching similar relative abundances as those found in the full-scale sewer biofilm. Concerning archaeal communities, no archaea other than methanogens were identified in pyrotag libraries from biofilm samples. Specifically, archaeal sequences were affiliated with three main genera, Methanosphaera, Methanobrevibacter, and Methanosaeta. Whereas sequences affiliated with Methanosphaera (relative abundances ranging from 10 to 23%) and Methanobrevibacter (76 to 86%) were prevalent during the first weeks of biofilm development (Fig. 3B), the archaeal community in 1-year-old biofilms was dominated mainly by sequences affiliated with the genus Methanosaeta, which were also prevalent in the biofilm collected from the full-scale sewer (Fig. 3B).

FIG 3.

Relative abundances of sequences (percent) affiliated with main bacterial classes (A) and main archaeal genera (B) in week 1, week 5, week 13, 1-year, and full-scale sewer biofilm samples.

Grouping of sequences into OTUs (97% cutoff) resulted in 1,283 and 137 OTUs for Bacteria and Archaea, respectively (see Table S5 in the supplemental material). OTU delineation allowed us to identify potentially those OTUs (i.e., species) that may make a relevant contribution to the development and activity of sulfidogenic and methanogenic biofilms. Because of the high diversity of the sample and nutrient availability in the system, OTUs were considered relevant in terms of abundance if their relative abundance was ≥4% in at least one sample. Whereas the relative abundance of some OTUs increased only at the end of the incubation period (OTU-B1, OTU-B6, and OTU-B7), that of others clearly decreased during this time (OTU-B3, OTU-B8, OTU-B10, OTU-B12, and OTU-B20) (Fig. 4A). One of the most prevalent OTUs in early stages of biofilm development (OTU-B3; >10% of total sequences) showed 100% sequence identity to Macellibacteroides fermentans, a fermentative member of the Porphyromonadaceae (Bacteroidetes) (49). Other common OTUs identified during this period (e.g., OTU-B8 and OTU-B20) were rare in mature and full-scale sewer biofilms. In turn, most prevalent OTUs in full-scale sewer biofilms were rare in the laboratory system, with the exception of OTU-B1 (83% sequence identity to Rikenella microfusus strain Q-1, an obligate anaerobic fermentative microorganism) (50). The bacterial community in the biofilm collected from the full-scale sewer was composed mainly of microorganisms affiliated with the class Betaproteobacteria (OTU-B2, OTU-B14, and OTU-B18) and the phyla Synergistetes (OTU-B4, OTU-B5, and OTU-B13) and Chloroflexi (OTU-B9) (Fig. 4A). Only OTU-B6 was affiliated with the class Deltaproteobacteria, having 99% sequence identity to Desulfobacter postgatei strain 2ac9.

FIG 4.

Bubble plots of bacterial (A) and archaeal (B) OTUs showing their relative abundances across samples, their taxonomy affiliation (at the genus level), and the percent identity to the first BLAST hit against reference sequence databases. Data are proportional to the radius and plotted on a logarithmic scale, as indicated below the graph. The relative abundance (percent) of each OTU at different sampling points is indicated next to the corresponding bubble (gray figures).

In order to study the phylogenetic structure of the SRB community during biofilm development in more detail, sequences affiliated with the class Deltaproteobacteria, which includes most of the sulfate reducers known to date, were retrieved and grouped into OTUs that were then used to construct a phylogenetic tree (see Fig. S9A in the supplemental material). Whereas abundant OTUs in the first weeks of incubation (OTU-D3 and OTU-D4) were phylogenetically related to Desulfobulbus propionicus strain DSM2032 (see Fig. S10 in the supplemental material), the composition of the SRB community changed as the biofilm developed. After 1 year of operation, the community was dominated mainly by OTU-D1 (36% of total deltaproteobacterial sequences), which showed 99% sequence identity to Desulfobacter postgatei strain 2ac9 (see Fig. S9A and Table S6 in the supplemental material). Although this OTU was also present in biofilms collected from a full-scale sewer, the deltaproteobacterial community under natural conditions was more diverse than that grown under laboratory conditions.

In turn, abundant archaeal OTUs (>4% of total sequences) were all affiliated with methanogenic lineages. Particularly, OTU-A1, which showed 99% sequence similarity to Methanosaeta concilii, was detected only in mature biofilms and in biofilms from the full-scale sewer (Fig. 4B; see also Fig. S9B in the supplemental material). In turn, OTU-A2 and OTU-A3 were detected mainly during the first weeks of biofilm growth. Both OTUs had 100% sequence identity to Methanobrevibacter smithii and Methanosphaera stadtmanae, respectively. Finally, OTU-A4 (showing 99% sequence similarity to Methanobrevibacter acididurans) was detected at low relative abundances in all pyrotag libraries analyzed.

Richness and diversity metrics calculated for the bacterial biofilm communities increased during the experimental period (see Table S7 in the supplemental material). However, the bacterial community in the biofilm from the full-scale sewer was less rich and diverse than that from biofilms under laboratory conditions. In turn, the richness of the archaeal community showed an opposite trend, clearly decreasing during the 13 weeks of incubation, but remained at a similar level in mature biofilms (see Table S7 in the supplemental material). Despite these changes in richness, archaeal diversity remained fairly constant from the start-up to the end of the monitoring period and decreased in mature biofilms. Moreover, both the richness and diversity of archaeal biofilm communities in the full-scale in-sewer biofilm were higher than the levels estimated for biofilms after 1 year of operation under laboratory conditions.

To easily compare bacterial and archaeal biofilm communities, samples were distributed in a nMDS two-dimensional (2D) ordination space according to their similarity based on the weighted UniFrac distance (see Fig. S11 in the supplemental material). The ordination segregated biofilm samples collected at early stages of development (weeks 1, 5, and 13) from those collected at mature stages from the laboratory-scale sewer and from the biofilm samples from the full-scale sewer. It is noteworthy that bacterial and archaeal communities in mature biofilms (i.e., 1 year of incubation) were similar to those occurring in biofilms from full-scale sewers.

DISCUSSION

Sulfide and methane production rates during biofilm formation.

In this study, we investigated the association between H₂S and CH₄ production and the corresponding biofilm development stage in a laboratory-scale anaerobic sewer pilot plant. H₂S production rates suggested a fully adapted and functional SRB community after 2 weeks of biofilm colonization. The low level of production of H₂S for the first 2 weeks may have been a consequence of the low abundance of SRB in young biofilms after the experimental setup (Fig. 4; see also Fig. S8 in the supplemental material). In turn, the higher level of H₂S production in R1 than in R2 and R3 may have resulted from the system design, considering that the bioreactors were connected in series and that wastewater that entered R2 and R3 contained only trace amounts of sulfate because of its consumption in R1.

Methane production rates measured in batch tests were minimal for the first 12 weeks, probably because reactors were filled with fresh wastewater (containing high concentrations of sulfate) just before the start-up of each batch test. The differences in CH₄ production and emission rates might be a consequence of biofilm adaptation under each reactor condition, which varied mainly in terms of the sulfate concentration and HRT. During normal functioning, the small quantity of sulfate in the R1 effluent could have promoted active methanogenesis in R2 and R3, whereas conditions in R1 (high sulfate and organic matter concentrations), in turn, favored SRB over MA (25, 51). Results from 6-h batch test experiments confirmed a stimulation of CH₄ production after 3 to 4 h of wastewater retention in the system (when sulfate was depleted), especially in R3, where the sulfate concentration was already low (see Fig. S4 in the supplemental material). These results point to a spatial segregation of microbial communities responsible for H₂S and CH₄ production along the length of the anaerobic sewer, although no direct evidences of this differential distribution were obtained. Further work is then needed to validate if both the composition and activity of SRB and MA communities in sewer biofilms vary along the length of full-scale sewer systems.

Sulfide and methane emissions by mature biofilms.

Comparison of H₂S emissions from young biofilms and those from mature biofilms showed a decrease as a consequence of the smaller amount of sulfate available in the influent wastewater. Notwithstanding this, the relative amount of sulfate reduced to H₂S increased in mature biofilms (from ≈80% to 100%). Concerning CH₄ emission, several factors could account for its increase in mature biofilms (from 17.9 ± 15.9 to 327.6 ± 16.6 mg COD-CH₄ liter⁻¹ day⁻¹), namely (i) the low sulfate concentration in the inlet wastewater after 1 year of the experiment favoring a higher methanogenic activity, (ii) the high rate of consumption of sulfate by SRB in mature biofilms stimulating CH₄ production, or (iii) a change in the composition of the methanogenic community over time toward species more adapted to local conditions, resulting in a higher level of production of CH₄.

Compositional changes of microbial communities.

DGGE fingerprints showed differences in the overall compositions of bacterial and archaeal communities between inlet wastewater and biofilm samples. Despite the inherent limitations of the PCR-DGGE approach (52), similarity analysis of both bacterial and archaeal communities based on DGGE band patterns grouped samples according to sampling date (i.e., stage of biofilm development), showing that the structure of microbial biofilm communities progressively adapted to local conditions in the system. The fact that both bacterial and archaeal communities showed similar clustering patterns suggests potential interactions (e.g., synergy or competition) that deserve further investigation.

During the first weeks of biofilm development, the most abundant OTUs belonging to the class Deltaproteobacteria (OTU-D3 and OTU-D4) were closely related to Desulfobulbus propionicus. Interestingly, this species was recently identified by Sun and coworkers as the main SRB in the outer layers of sewer biofilms (10). D. propionicus reduces sulfate via the incomplete oxidation of organic acids such as lactate, propionate, butyrate, and ethanol to acetate (53), all of which were available in the inlet wastewater. In turn, the SRB community in mature biofilms was composed mainly of a deltaproteobacterium closely related to Desulfobacter postgatei (OTU-D1), whereas sequences affiliated with SRB colonizers (i.e., OTU-D3 and OTU-D4) were rare after 1 year of incubation (see Fig. S10 in the supplemental material).

Hydrogenotrophic methanogens (belonging to the order Methanomicrobiales or Methanobacteriales) may use H₂ generated in fermentative metabolisms or act as hydrogen scavengers in syntrophic growth with acetate-oxidizing microorganisms (54–57). Also, acetate produced during fermentation of organic substrates by anaerobic heterotrophs within the biofilm matrix would be used by acetoclastic methanogens (Methanosarcinaceae and Methanosaetaceae) (58). The identification of sequences belonging to both groups of methanogens (hydrogenotrophic and acetoclastic) in our experimental system during the study period lends support to a progressive change of methanogenic pathways over time in relation to both local environmental conditions and the composition of the archaeal community at each stage of biofilm development.

Methanobrevibacter smithii and Methanosphaera stadtmanae (Methanobacteriales) are considered to be the prevalent methanogens in the human gut (59). In our study, sequences belonging to both species were identified in DGGE fingerprints from inlet wastewater samples and in pyrotag libraries from the first weeks of biofilm development, suggesting that archaeal colonizers at early stages of biofilm development derive from human fecal material in wastewater. These human-derived methanogens were probably outcompeted later on by acetoclastic methanogens (e.g., Methanosaeta concilii), which would probably be more adapted to environmental conditions in the pilot plant. The time needed by these better-adapted methanogens to be established in the biofilm matrix is consistent with the low level of CH₄ production during the initial phases of biofilm development. During this first stage, methanogenesis was also probably inhibited by sulfate reducers, which decrease the H₂ potential pressure below levels required by methanogens when sulfate is not limiting (60). Despite the well-known competitive interaction between SRB and MA, several studies have demonstrated that both groups coexist under certain conditions (60, 61). Particularly, Struchtemeyer and coworkers reported that low levels of sulfate may favor acetate consumption by MA rather than by SRB (62). In this regard, and although it is always risky to infer functional properties from phylogeny (63), sequences affiliated with both Deltaproteobacteria and MA identified in mature biofilms were closely related to species that are able to use acetate (i.e., D. postgatei and M. concilii, respectively). Accordingly, the increase in CH₄ production measured after 1 year of incubation might be explained by the establishment of acetoclastic methanogens in the biofilm, favored by a greater availability of acetate in wastewater. Besides, the increase in CH₄ production also could have been favored by the stratification of both groups within the biofilm matrix, as recently reported (10), although in our case, no measurements aimed at resolving the spatial organization of SRB and MA in the studied biofilms were carried out.

Altogether, this study provides data that confirm the capacity of our laboratory experimental system to mimic the functioning of full-scale sewers both microbiologically and operationally in terms of H₂S and CH₄ production and the composition of microbial communities during biofilm growth. Whereas H₂S emission was notably high during early stages of biofilm development, CH₄ emissions increased after biofilm maturation, coinciding with an establishment of a methanogenic community better adapted to sewer conditions; for this reason, it is important to take into account that the management of sewer systems is very important from the first stages of sewer functioning. Although further research is needed to better resolve the dynamics of the bacterial communities in biofilms and to identify the key bacterial players involved in both nutrient transformations and potential syntrophic interactions that occur in these complex ecosystems, our results should be valuable when designing optimal strategies to mitigate H₂S and CH₄ emissions from sewer systems.