Moustapha
Harb‡
a,
Yanghui
Xiong‡
a,
Jeremy
Guest
b,
Gary
Amy
a and
Pei-Ying
Hong
*a
aWater Desalination and Reuse Center, Environmental Science and Engineering, King Abdullah University of Science and Technology (KAUST), 4700 King Abdullah Boulevard, Thuwal 23955-6900, Saudi Arabia. E-mail: peiying.hong@kaust.edu.sa; Tel: +966 12 8082218
bDepartment of Civil and Environmental Engineering, University of Illinois at Urbana-Champaign, Urbana, Illinois 61801, USA
First published on 14th August 2015
Two lab-scale anaerobic membrane bioreactors (AnMBRs), one up-flow attached-growth (UA) and another continuously stirred (CSTR), were operated under mesophilic conditions (35 °C) while treating synthetic municipal wastewater (800 mg L−1 COD). Each reactor was attached to both polyvinylidene fluoride (PVDF) and polyethersulfone (PES) microfiltration (MF) membranes in an external cross-flow configuration. Both reactors were started up and run under the same operating conditions for multiple steady-state experiments. Chemical oxygen demand (COD) removal rates were similar for both reactors (90–96%), but captured methane was found to be 11–18% higher for the CSTR than the UA reactor. Ion Torrent sequencing targeting 16S rRNA genes showed that several operational taxonomic units (OTUs) most closely related to fermentative bacteria (e.g., Microbacter margulisiae) were dominant in the suspended biomass of the CSTR, accounting for 30% of the microbial community. Conversely, methanogenic archaea (e.g., Methanosaeta) and syntrophic bacteria (e.g., Smithella propionica) were found in significantly higher relative abundances in the UA AnMBR as compared to the CSTR due to their affinity for surface attachment. Of the methanogens that were present in the CSTR sludge, hydrogenotrophic methanogens dominated (e.g., Methanobacterium). Measured EPS (both proteins and carbohydrates), which has been broadly linked to fouling, was determined to be consistently lower in the UA AnMBR membrane samples than in CSTR AnMBR membrane samples. Principal component analysis (PCA) based on HPLC profiles of soluble microbial products (SMPs) further demonstrated these differences between reactor types in replicate runs. The results of this study showed that reactor configuration can significantly impact the development of the microbial communities of AnMBRs that are responsible for both membrane and reactor performance.
Water impactAnaerobic membrane bioreactors (AnMBRs) have recently gained consideration as an alternative technology for the treatment of municipal wastewater. Despite that, the effect of reactor configurations on their performance and sustainability remains understudied. This work aims to evaluate two different AnMBR types at low organic-loading conditions. Results show that the microbial consortium developed in each reactor can potentially have a significant impact on both biogas production and the microbial products that affect biofouling. |
Nonetheless, there are a number of limitations in the operation of AnMBRs. One major issue that is directly related to AnMBR sustainability is that of methane recovery capacity. It is still not known to a high degree of accuracy how much energy can be captured per unit of chemical oxygen demand (COD), mainly due to the widely variable operational conditions reported in existing AnMBR literature.4 At high organic loading rates, methane in biogas can approach the maximum theoretically recoverable.5 Although lower operational temperatures increase methane fractions lost in AnMBR effluent,6 low HRTs and organic loading conditions have been shown to significantly reduce methane captured in produced biogas even at mesophilic temperatures.7 With this in mind, methane recovery efficiency remains one of the primary obstacles when evaluating AnMBRs alongside other treatment technologies for municipal wastewater.8,9 Several anaerobic reactor types have been studied in AnMBRs along with their biogas production rates.10 However, the effect of the microbial community differences between these reactor types on methane production is not well understood.
The most dominant obstacle preventing full-scale implementation of AnMBRs remains the issue of membrane biofouling and its impact on the overall sustainability of the AnMBR.2,4 Biofouling control processes are often energy intensive and, as a result, can effectively negate the advantages of using anaerobic digestion.9 This is especially problematic when applying AnMBR technology to municipal and other low-strength wastewaters. The high degradation rates and low biomass production associated with anaerobic digestion have the effect of requiring a higher organic loading rate (OLR) for process stability than can be achieved with a municipal wastewater at normal hydraulic retention times (HRTs).11 Reducing HRT reduces the fraction of the soluble microbial products (SMPs) that are degraded, which accelerates the effects of membrane biofouling.12 The impact of reactor configuration on overall performance and membrane fouling in AnMBRs has been studied,13–15 showing that attached growth reactors have the potential to reduce the effects of biofouling. Furthermore, it has been established that SMP generation is a significant contributor to biofouling.3,16 Little is known, however, about how microbial community differences between reactor types affect SMP generation and biofouling in AnMBRs.
Despite a number of articles addressing AnMBR operation and membrane fouling,14,17–19 the field still lacks a thorough understanding of how microbial community structure and interactions influence the reactor and membrane performance characteristics related to process sustainability.20–22 Our study evaluated different reactor configurations in AnMBRs from a microbial community perspective to improve this understanding. We hypothesized that each reactor configuration would select for different bacterial populations in the sludge and would result in differences in anaerobic treatment efficiency. It was further hypothesized that the differences in reactor configurations would also lead to differences in the generated SMPs and membrane fouling rates. The focus of this work is on studying the microbial communities in different AnMBR reactor configurations (i.e., CSTR and UA) while relating their dynamics to reactor performance. This work also examines SMPs in the suspended sludge of each reactor type and the extracellular polymeric substance (EPS) content of the fouled membrane biofilms, which are critical components governing the sustainability and net energy consumption of AnMBR systems.
Fig. 1 Configurations and operating conditions for up-flow attached-growth (UA) and continuously stirred (CSTR) anaerobic MBRs studied. |
Biomass regeneration rates and losses in the effluent were calculated to estimate actual methane yield per unit COD. A microorganism decay rate of 0.03 d−1 and COD conversion coefficient of 1.42 g COD per g microorganism were assumed.25 To determine CO2 and CH4 lost in the effluent, batch tests to manually strip all dissolved gasses were performed after cessation of methane production. The dissolved biogas susceptible to being lost in the effluent was in the liquid phase of both the UA and the CSTR AnMBRs, as it was the only portion of sludge that was in contact with the membranes and subject to filtration. As such, suspended sludge was used for both reactors. Tests were done in triplicate on 10 mL sludge samples in 30 mL vials from both reactors. In order to strip the dissolved gasses completely, the headspace-to-liquid ratio was increased to approximately 30 times that of the reactors and filled with 100% nitrogen to increase the batch tests' mass transfer capacity. Test vials were vortexed for 5 min and then placed on a shaker at 100 rpm for 24 h at 35 °C. The 20 mL headspace was then measured for methane content. Methane yield was calculated based on stoichiometric conversion rates of proteins, lipids, and carbohydrates and their ratios in the synthetic wastewater.25
For analysis of the microbial community in the attached membrane biofilms, the biomass was divided into two portions: (i) loosely bound and (ii) strongly bound. For the loosely bound biofilm portion, three freshly harvested 4 cm2 membrane segments, cut from sections equally spaced along the length of the membrane surface, were each dipped lightly five times in 6 mL of 1× PBS. The suspension was then divided into 2 mL aliquots and pelleted by centrifuging for 30 min at 9400g and stored at −20 °C. For the strongly bound biofilm portion, each of the same three separate segments was aseptically cut into strips and placed into a 2 mL tube with 2 mL of 1× PBS. The tubes were vortexed for 5 min and then sonicated for 5 min at 25% frequency and 2 s pulsating intervals using the QSonica Q500 Sonicator (QSonica LLC). The suspension was then pelleted by centrifuging for 30 min at 9400g and stored at −20 °C.
The first approach assigned sequences to bacterial/archaeal taxonomic hierarchy at a 95% confidence level using the Ribosomal Database Project (RDP) Classifier.29 The relative abundances of the bacterial and archaeal genera were calculated, collated and then square-root transformed. The transformed data set were then computed for their Bray–Curtis similarities and represented graphically for spatial distribution in a bootstrapped metric multidimensional scaling (mMDS) plot using Primer-E version 7.30 One-way analysis of similarity (ANOSIM) was performed as a test statistic using Primer-E version 7 to evaluate differences among clustering patterns of the mMDS plot. One-way ANOSIM generates a denominator constant R, ranging from 0 to 1 with a corresponding P value. The R statistic provides a comparative measure of the degree of separation among clustered patterns with R = 0 representing no differences between compared sample groups and R = 1 representing sample groups with complete dissimilarity.31
In the second approach, sequence files were sorted based on ratio of identical sequences to total sequences and identified as unique operational taxonomic units (OTUs). All chimera-removed sequence files were combined with an in-house written Perl script. The combined sequence file was then sorted for unique OTUs at 97% 16S rRNA gene similarity using CD-Hit.32 OTUs were prioritized based on highest average relative abundance in each reactor and sorted accordingly. OTUs with an average relative abundance in either the UA or the CSTR AnMBR of above 0.3% of total sequences were blasted against the NCBI nucleotide database using the BLASTN algorithm to check for their closest matching bacteria or archaea species. Results of sequence blasting were then recorded for NCBI sequence identity similarity data (closest classified species, identity match, maximum score, and E-value). For the comparison of average relative abundances of sample sets, an independent (unpaired) two-tailed t-test was used with an assumption of unequal variance between sample sets.
All high-throughput sequencing files were deposited in the Short Read Archive (SRA) of the European Nucleotide Archive (ENA) under study accession number PRJEB8654.
Analysis of the molecular weight of SMPs and soluble EPS was conducted with a Waters Breeze™ 2 HPLC System (Waters Chromatography, Milford, MA, USA) equipped with a Binary HPLC pump (Waters 1525), Auto sampler (Waters 2707) and UV/Visible Detector (Waters 2489). The SEC method was carried out with a Shodex KW-802.5 column. The injection volume for each sample was 20 μL with 20 min retention time, and the mobile phase consisted of 50 mM phosphate buffer (pH = 7.0) with 300 mM NaCl. The detection was carried out at room temperature with a UV detector at 210 and 280 nm wavelengths, which are commonly used for EPS.35 UV absorbance at both wavelengths has been used extensively for protein determination.36 Based on previous studies that have used UV for excitation, 210 nm wavelengths represent protein-like substances while 280 nm wavelengths more commonly detect humic-like substances.37,38 Molecular weights of compounds detected were calculated based on a calibration curve (y = −0.454x + 8.38) established for proteins of known molecular weights (x) and their corresponding peak retention times (y).
The hydrophobicity of each membrane was assessed using contact angles, which were determined by the static drop technique with a KSV Cam 200 optical contact angle meter (KSV Instruments Ltd., Finland). Briefly, deionized water was made to come into contact with the membrane surface by a micro-syringe to a droplet size of diameter 0.4–0.5 cm. At least three different contact angle measurements were carried out for each piece of membrane to obtain an average.
The surface charge of each membrane was determined by zeta potential, which was measured using a Surpass Complete electrokinetic analyzer (Anton Paar, GmbH, Graz, Austria). The membranes were equilibrated with the electrolyte (0.5 M NaCl) for at least 30 min before measurement. The electrolyte was titrated stepwise from pH 3 to 10 in increments of 0.5 with triplicate measurements taken at each step. The zeta potential of the membrane was automatically calculated by the analyzer from the streaming potentials using the Helmholtz–Smoluchowski equation with the Fairbrother and Mastin substitution to correct for membrane surface conductance.39
UA AnMBR | CSTR AnMBR | Max. th. | |||
---|---|---|---|---|---|
Run 1 (n = 15) | Run 2 (n = 21) | Run 1 (n = 19) | Run 2 (n = 27) | ||
a Total CH4 = CH4 captured + CH4 measured in effluent. b Based on available COD not converted to biomass. c Based on 33% electrical conversion efficiency. | |||||
Captured CH4 (mL d−1) | 225 ± 34 | 223 ± 40 | 263 ± 50 | 250 ± 52 | 342 |
Captured CH4 (mL g−1 COD) | 236 ± 35 | 234 ± 41 | 269 ± 51 | 256 ± 53 | 350 |
Total CH4a (mL g−1 COD) | 298 ± 38 | 296 ± 45 | 326 ± 55 | 313 ± 57 | 350 |
Captured potential energyb (kW h m−3 WW) | 1.25 ± 0.21 | 1.26 ± 0.25 | 1.43 ± 0.30 | 1.37 ± 0.31 | 1.90 |
Usable potential energyc (kW h m−3 WW) | 0.42 ± 0.07 | 0.42 ± 0.08 | 0.48 ± 0.10 | 0.46 ± 0.10 | 0.63 |
Fig. 3 Metric multi-dimensional scaling (mMDS) plot of microbial community similarities for all samples from up-flow attached-growth (UA) and continuously stirred (CSTR) AnMBRs for runs 1 and 2. |
(A) Closest species match | UA avg. (%) | CSTR avg. (%) | UA/CSTR ratio | P-value | Identity match (%) |
---|---|---|---|---|---|
Run 1 | |||||
Smithella propionica | 1.39 | 0.163 | 8.51 | <0.0001 | 98 |
Levilinea saccharolytica | 0.573 | 0.023 | 24.3 | 0.006 | 98 |
Methanobacterium formicicum | 0.137 | 1.74 | 0.079 | 0.009 | 98 |
Aminivibrio pyruvatiphilus | 0.507 | 0.223 | 2.27 | 0.004 | 98 |
Geobacter lovleyi | 2.31 | 0.179 | 12.9 | 0.036 | 98 |
Run 2 | |||||
Smithella propionica | 2.83 | 0.087 | 32.6 | <0.001 | 95 |
Levilinea saccharolytica | 0.513 | 0.033 | 15.5 | <0.0001 | 98 |
Methanobacterium formicicum | 0.110 | 0.416 | 0.24 | 0.013 | 96 |
Methanospirillum stamsii | 0.863 | 0.005 | 185 | 0.007 | 97 |
Methanosaeta concilii | 1.11 | 0.149 | 7.45 | 0.005 | 99 |
(B) Closest species match | UA sludge avg. (%) | CSTR sludge avg. (%) | UA/CSTR sludge ratio | P-value | Identity match (%) |
---|---|---|---|---|---|
Run 1 | |||||
Microbacter margulisiae | 0.016 | 20.4 | 7.6 × 10−4 | 0.002 | 85 |
Lutispora thermophila | 2.90 | 10.1 | 0.286 | <0.001 | 88 |
Run 2 | |||||
Prolixibacter bellariivorans | 0.0007 | 23.8 | 2.95 × 10−5 | <0.001 | 86 |
In contrast, one methanogenic species was determined to be more prevalent in the CSTR AnMBR. Methanobacterium formicicum was present in the CSTR at a higher relative abundance than in the UA for both runs 1 and 2 with UA/CSTR RA ratios of 0.079 and 0.24, respectively (unpaired t-test, P = 0.009 and 0.013). Additionally, several OTUs were identified that made up a significant overall percentage of the CSTR sludge (20–30%) throughout both runs 1 and 2 (Table 2B). OTUs closest related to Microbacter margulisiae and Lutispora thermophila made up an average relative abundance of 20.4% and 10.1%, respectively, in the CSTR sludge during run 1. In run 2, two OTUs most closely related to Prolixibacter bellariivorans made up an average relative abundance of 23.8% of the CSTR sludge. All three of these OTUs had significantly lower representation in the UA reactor sludge with UA/CSTR sludge RAs of 7.61 × 10−4, 0.286, and 2.95 × 10−4, respectively (unpaired t-test, P = 0.002, P < 0.001 and P < 0.001).
Similarly, genus-based analysis revealed that key bacterial and archaeal genera were present in the attached biomass of the UA at significantly higher relative abundances than in the suspended biomass of both the UA and the CSTR. To illustrate, Syntrophobacter was present in the UA at an attached to suspended sludge ratio of 6.32 in run 1 (unpaired t-test, P = 0.031) and 10.4 in run 2 (unpaired t-test, P = 0.0023). The UA-attached sludge to CSTR sludge (suspended) ratio for Syntrophobacter was also higher in both runs 1 and 2 at 4.07 and 1.48, respectively (unpaired t-test, P = 0.096 and P = 0.042). Several other genera identified as syntrophic bacteria were found to have UA-attached to UA-suspended sludge ratios and UA-attached to CSTR sludge ratios of greater than 1 (ranging from 2.45 to 127) with high significance (Table 3) in runs 1 and 2. Among these genera were Smithella, Syntrophomonas, and unclassified Syntrophobacteraceae. Other bacteria that favored attached sludge as compared to suspended were Geobacter and Desulfovibrio, with UA-attached to UA-suspended sludge ratios ranging from 6.43 to 8.32 (unpaired t-test, P < 0.05) and UA-attached to CSTR sludge ratios ranging from 2.43 to 32.2 (unpaired t-test, P < 0.07).
UA att./UA SS ratio | P-value | UA att./CSTR SS ratio | P-value | |
---|---|---|---|---|
a Normalized based on total relative abundance of archaea. | ||||
Run 1 | ||||
Syntrophomonas | 15.6 | 0.0013 | 3.89 | 0.061 |
Smithella | 2.85 | 0.018 | 6.69 | 0.058 |
Syntrophobacter | 6.32 | 0.031 | 4.07 | 0.096 |
Unclassified Syntrophobacteraceae | 11.8 | 0.0066 | 39.7 | 0.060 |
Desulfovibrio | 6.43 | 0.049 | 32.7 | 0.034 |
Methanobacterium | 1.05 | 0.23 | 0.778 | 0.034 |
Methanosaeta | 2.74 | 0.058 | 1.64 | 0.15 |
Run 2 | ||||
Smithella | 2.45 | 0.037 | 3.33 | 0.026 |
Syntrophobacter | 10.4 | 0.0023 | 1.48 | 0.042 |
Geobacter | 8.32 | 0.018 | 4.44 | 0.026 |
Desulfovibrio | 6.47 | 0.031 | 2.43 | 0.065 |
Unclassified Syntrophobacteraceae | 41.2 | 0.0030 | 127 | 0.0031 |
Methanobacterium | 0.299 | 0.0070 | 0.523 | <0.0001 |
Methanospirillum | 3.26 | 0.0098 | 44.1 | 0.0058 |
Methanosaeta | 4.51 | 0.0022 | 3.89 | 0.0024 |
Relative abundances of individual methanogens were compared on a percent of total archaeal sequences basis. These results showed that for run 1, Methanobacterium had the highest relative abundance among methanogens in the UA-attached, UA-suspended, and CSTR sludge biomass at 42.3 ± 13.9%, 40.3 ± 9.2%, and 54.4 ± 8.9%, respectively. For run 2, however, the UA-attached archaeal community was dominated by Methanosaeta with an average relative abundance of 43.8 ± 9.1% while UA-suspended and CSTR sludge biomass during run 2 remained dominated by Methanobacterium at 65.0 ± 13.9% and 37.1 ± 6.6%, respectively. UA-attached to UA-suspended ratios of Methanobacterium were 1.05 and 0.299 (unpaired t-test, P = 0.23 and P = 0.0070) while UA-attached to CSTR sludge ratios were 0.778 and 0.523 (unpaired t-test, P = 0.034 and P < 0.0001) for runs 1 and 2, respectively. For Methanosaeta, UA-attached to UA-suspended ratios were significantly higher at 2.74 and 4.51 (unpaired t-test, P = 0.058 and P = 0.0022) for runs 1 and 2, respectively. Similarly, UA-attached to CSTR sludge ratios for Methanosaeta were 1.64 and 3.89 (unpaired t-test, P = 0.15 and P = 0.0024).
Analysis of HPLC results for membrane biofilm EPS contents showed that protein profiles detected at 210 nm consistently contained the 600 kDa MW compound. The peak size of this compound increased for membranes that experienced premature increases in total membrane resistance (Fig. 7). Absorbance profiles at 210 nm for soluble membrane biofilm EPS showed a 3-fold or greater increase in peak height at this retention time between the CSTR PES membrane and the CSTR PVDF membrane during run 1 (Fig. 6C). Likewise, peak height of the 210 nm protein fragment was greater in the UA PES membrane as compared to the UA PVDF membrane in run 2 (Fig. 6B). This peak was also seen in absorbance profiles detected at 280 nm and followed a similar trend, although peak heights were consistently lower than those detected at 210 nm.
The CSTR consistently achieved higher methane content in the total biogas and produced a significantly higher daily methane yield compared to the UA AnMBR. The mass transfer limitations that were responsible for considerable methane losses in the AnMBR effluents were likely also affected by differences in the operational conditions of the CSTR as compared to the UA reactor. Given that the CSTR was operated under constant internal mixing conditions while the UA was not, it seems reasonable that slightly better transfer rates of methane from the CSTR sludge to the biogas could have accounted for the higher biogas and methane yields seen in the CSTR. Furthermore, in most cases studied for anaerobic digestion, hydrolysis and fermentation were determined to be rate-limiting steps.43 The dominance of up to 30.5% of the total microbial community within the CSTR sludge by three OTUs most closely related to the fermentative bacteria Microbacter margulisiae (run 1), Lutispora thermophila (run 1), and Prolixibacter bellariivorans (run 2) (Table 2B) suggests that the CSTR AnMBR in this study was likely able to overcome the hydrolysis and/or fermentation digestion steps more efficiently than the UA AnMBR system, possibly leading to higher overall methane yields. Although the specific fermentative bacterial species changed between runs 1 and 2, their prominence in the CSTR sludge and overall role in the anaerobic digestion process as fermenters remained consistent. During the time period between the end of run 1 and the commencement of run 2, both reactors were operated in batch mode, which could have led to these changes.
Another possible explanation for the lower methane production by the UA could be related to hydrogen- and/or acetate-utilizing bacteria competing with methanogens in the UA biomass. For example, an OTU closest related to Geobacter lovleyi was present at a higher relative abundance in the UA in both runs 1 and 2 (UA/CSTR ratio = 12.9, unpaired t-test, P = 0.036). Geobacter lovleyi is able to utilize both hydrogen and acetate under anaerobic conditions and could decrease the amount of both substrates that are taken up by the hydrogenotrophic and acetoclastic methanogens.44 This, in turn, would result in lower methane and higher carbon dioxide content within the biogas.
Other significant differences in the microbial populations of the UA and CSTR AnMBRs were also observed, specifically those associated with the attached and suspended biomass. Our findings have shown that methanogens and archaea in general have a more dominant presence in the attached biomass as compared to suspended sludge (average RA ratio = 8.44, unpaired t-test, P < 0.001), along with syntrophs and sulfate-reducing bacteria. The inherently longer sludge age of the attached biomass likely favors these microbial groups as compared to other bacteria due to their relatively slower growth rates.25
Different environmental conditions are likely to persist in attached biofilms due to the vertical gradient of substrates that may also drive niche differentiation of archaea or syntrophs over others, in turn providing these bacterial populations with a competitive edge than in the suspended form under lower substrate concentrations. To illustrate, an OTU identified as propionate-degrading bacterium Smithella propionica had a significantly higher RA in the UA AnMBR as compared to the CSTR AnMBR (Table 2A).46Methanospirillum stamsii is a methanogen known to grow in a syntrophic propionate-oxidizing anaerobic consortium.47 The relative abundance of this species showed significant correlation with the relative abundance of S. propionica (Spearman's rank correlation ρ = 0.53, P = 0.034). Although S. propionica produces acetate that could be utilized by acetoclastic methanogens, the relative abundance of S. propionica did not exhibit a significant correlation with Methanosaeta concilii (Spearman's rank correlation ρ = 0.35, P = 0.19), suggesting that the transfer of hydrogen was the more mutualistic interaction when compared to the transfer of acetate to acetoclastic methanogens. Hence, attachment or aggregation facilitates the interspecies transfer of electrons, hydrogen and/or formate, which is a key requirement in obligate syntrophic interactions,45 and is potentially responsible for the higher abundances of methanogens and syntrophs in the attached biomass.
The transfer of hydrogen to methanogens can, however, be limited by diffusion when sulfate-reducing bacteria are present in the attached biomass. It has been previously shown that the transfer of electron carriers between cells is only limited by diffusion rates when cells are within 2–3 μm of each other (i.e. aggregated biomass), while H2-producing and/or utilizing microbes do not affect inter-cell transfer rates when at distances >10 μm (i.e. suspended biomass) except in their contribution to the bulk solution.48 With this in mind, the presence of Desulfovibrio at significantly higher relative abundances in the attached biomass (Table 3) likely created less favorable conditions for hydrogenotrophic methanogens in the attached biomass than in the suspension.
The majority of archaea in the CSTR AnMBR were composed of hydrogenotrophic methanogens, specifically Methanobacterium formicicum, reaching up to 80% of the total archaeal community. Hydrogenotrophic methanogenesis has been previously identified as the dominant methane-producing pathway in anaerobic digesters.49 As such, the higher abundance of hydrogenotrophic methanogens as compared to their acetoclastic counterparts may have contributed to higher carbon dioxide utilization and ultimately higher methane production by the CSTR AnMBR. It was observed in a previous study that increases in shear rates within an AnMBR coincided with increases in the hydrogenotrophic methanogenic population of the reactor.41 This is consistent with our findings of hydrogenotrophic methanogen dominance in the CSTR, as shear rates in the CSTR were likely inherently higher compared to the UA reactor given that recirculation rates in the two reactors were the same, but the CSTR was also rigorously internally mixed by an impeller.
Existing research has shown that high molecular weight compounds can dominate proteinaceous EPS and SMPs in MBRs.14,50 Furthermore, a previous study comparing reactor configurations of AnMBRs found that there were significant differences in SMP molecular weight fractionation between suspended growth and granular AnMBRs.15 Our findings suggest that differences in the bacterial populations between reactor configurations have also likely accounted for differences in the SMP constituents of CSTR and UA retentates (Fig. 5). Major differences in total membrane resistance were observed between run 1 and run 2 for the PES membrane in the CSTR. This was likely due to differences in the proteinaceous SMP characteristics, as they varied more in the CSTR as compared to the UA reactor. To illustrate, a peak representing a 100 kDa protein-like compound was only present in select samples of retentate and permeate from the CSTR in run 1 (Fig. 6). The timing of this peak's presence coincided with increases in total resistance in the CSTR PES membrane at days 15 and 36 of run 1. This compound may have played a role in membrane fouling by pore adsorption and accumulation given its relatively small molecular weight compared to the membrane pore size. Conversely, this peak was not detected in CSTR SMP samples from run 2 or in any UA SMP samples, even though the UA PES membrane in run 2 also experienced premature increases in total resistance.
Previous research suggests that membrane performance in MBRs is negatively affected by the concentration of colloidal matter in reactor suspension and, as a result, favors attached growth type AnMBR systems.13,15 Our results showed that despite differences in the SMP constituents, membrane fouling was minimal for both reactors when the transmembrane fluxes were maintained below 8 L m−2 h−1, with the exception of the CSTR PES membrane in run 1 and, to a lesser degree, the UA PES membrane in run 2. Proteins and carbohydrates, which have been previously found to influence fouling and adversely impact membrane performance,21,51,52 were higher in CSTR PES membrane biofilms than in corresponding PVDF biofilms. Most of the characterized membrane properties showed no significant difference between virgin PES and PVDF membranes (Table S1†). However, the PES membrane used in this study exhibited a slightly positive zeta potential, which may have resulted in an enhanced affinity for the negatively charged microbial cells. In a study by Gao et al., different membrane types (i.e., polyether block amide-coated PVDF and uncoated polyetherimide) were tested and their findings revealed that uncoated polyetherimide had a greater surface affinity for Bacteroidetes.21 This, in turn, led to a predominantly proteinaceous EPS layer and a faster fouling rate than the coated PVDF membranes. This observation coincided with our findings that both the fouled PES membrane of the CSTR for run 1 and the fouled PES membrane of the UA for run 2 had 3-fold or greater increases in peak height for a 600 kDa protein fragment detected on HPLC as compared to PVDF membranes in those same runs (Fig. 7).
Normalized total average protein and carbohydrate contents were higher for CSTR PES membranes in both runs 1 and 2 (Fig. 4) even though the PES membrane in the CSTR during run 2 did not exhibit the same fouling rates as in run 1. Additionally, the lower average protein and carbohydrate content seen in PES biofilms in the UA AnMBR suggests that reactor configuration might also play a role in EPS affinity for a specific membrane type. In future studies, it would be useful to isolate and identify the observed peaks representing the 600 kDa (EPS and SMPs) and 100 kDa (SMPs) compounds in the HPLC profiles in order to determine if a link can be established between those compounds and the bacteria present in the reactor sludge and membrane biofilms.
Considering that the treatment efficiencies of both reactors were consistently similar in terms of COD removal rates, the main performance characteristics potentially affected by reactor type are methane production capacity and biofouling rates. Both of these parameters directly impact process efficiency and sustainability, which are the primary obstacles limiting the application of AnMBRs for municipal wastewater treatment. Through the use of molecular tools, the links between fundamental microbiology and the performance of these systems were examined, providing insights useful for future design and operation considerations.
Microbial community analysis shows that methanogens, syntrophs, and fermentative and sulfate-reducing bacteria are all significantly affected by differences in AnMBR configuration. Results suggest that these differences can lead to variations in biogas production and microbe-produced foulant compositions, both of which are integral to the sustainability of AnMBR application to municipal wastewater. Furthermore, although membrane characteristics appear to have an impact on biofilm formation, reactor stability and microbial foulant composition may contribute more directly to membrane fouling. These findings illustrate the importance of understanding microbial community interactions when evaluating AnMBRs in future research for wastewater treatment applications.
Footnotes |
† Electronic supplementary information (ESI) available. See DOI: 10.1039/c5ew00162e |
‡ Equal contributing authors. |
This journal is © The Royal Society of Chemistry 2015 |