Research paper

Two-Phase Improves Performance of Anaerobic Membrane Bioreactor Treatment of Food Waste at High Organic Loading Rates Yamrot M. Amha,† Michael Corbett,‡ and Adam L. Smith*,†

†Astani Department of Civil and Environmental Engineering, University of Southern California, 3620 South Vermont Avenue, Los Angeles, California 90089, United States ‡Divert, Inc., 23 Bradford Street, 3rd Floor, Concord, Massachusetts 01742, United States

*S Supporting Information

ABSTRACT: Anaerobic membrane bioreactors (AnMBRs) are in use at the full-scale for energy recovery from food waste (FW). In this study, the potential for two-phase (acid/gas) AnMBR treatment of FW was investigated as a strategy to increase microbial diversity, thereby improving performance. Two bench-scale AnMBRs were operated in single-phase (SP) and two-phase (TP) mode across incremental increases in organic loading rate (OLR) from 2.5 to 15 g total chemical oxygen demand (COD) L·d−1. The TP acid-phase (TP-AP) enriched total VFAs by 3-fold compared to influent FW and harbored a distinct microbial community enriched in fermenters that thrived in the low pH environment. The TP methane phase (TP-MP) showed increased methane production and resilience relative to SP as OLR increased from 3.5 to 10 g COD L·d−1. SP showed signs of inhibition (i.e., rapid decrease in methane production per OLR) at 10 g COD L·d−1, whereas both systems were inhibited at 15 g COD L·d−1. At 10 g COD L· d−1, where the highest difference in performance was observed (20.3% increase in methane production), activity of syntrophic bacteria in TP-MP was double that of SP. Our results indicate that AnMBRs in TP mode could effectively treat FW at OLRs up to 10 g COD·L day−1 by improving hydrolysis rates, microbial diversity, and syntroph activity, and enriching resistant communities to high OLRs relative to AnMBRs in SP mode.

1. INTRODUCTION

As landfills rapidly reach capacity in the US and elsewhere, diversion of organic wastes is expected to become the norm. Anaerobic membrane bioreactors (AnMBRs), which combine anaerobic treatment with membrane separation, have emerged as a sustainable food waste (FW) management strategy with reduced environmental footprint relative to landfilling and composting, while also providing energy recovery via biogas production.1 Compared to conventional anaerobic digesters (ADs), membrane separation in AnMBRs decouples solid retention time (SRT) and hydraulic retention time (HRT), enabling operation at longer SRTs. This can be advantageous for FW management due to the high organic content and temporal heterogeneity in waste characteristics. The long SRT and membrane separation drastically improves effluent quality relative to conventional AD, an important feature in decentralized FW management where effluent discharge to local publicly owned treatment works is necessary. AnMBRs may also permit operation at higher OLR than ADs, a critical parameter that dictates system capacity and reactor dimen- sions. Optimization of performance of decentralized FW management strategies, such as AnMBRs, is needed to increase treatment capacity, reduce environmental impacts, and improve economic benefits of resource recovery.

Anaerobic treatment of FW can be accomplished in single- phase (SP) or two-phase (TP) systems. TP anaerobic digestion separates the microbial conversion to an acid-phase (TP-AP) and methane-phase (TP-MP) digester operated in series. The TP-AP provides a unique biochemical environment supporting the growth of fermentative bacteria, while methanogenic archaea are suppressed by the low pH and short SRT (typically less than 3 days). The TP-AP digester also acts as an equalization tank, protecting the TP-MP from pulses in organic loading or potential inhibitors that could result in performance instability.2 Relative to anaerobic digestion of wastewater sludges, decentralized FW manage- ment systems experience greater temporal variability in feedstock quality and strength. Therefore, TP could provide greater benefit in this waste management scenario relative to SP. The TP-MP digester is operated near neutral pH at a longer SRT to enrich for methanogenic archaea and other microbial populations with low growth rates. Overall, TP has been reported to allow for higher OLRs and can improve biogas production and quality relative to SP.3,4

Received: May 1, 2019 Revised: July 26, 2019 Accepted: July 29, 2019 Published: July 29, 2019

Article

pubs.acs.org/estCite This: Environ. Sci. Technol. 2019, 53, 9572−9583

© 2019 American Chemical Society 9572 DOI: 10.1021/acs.est.9b02639 Environ. Sci. Technol. 2019, 53, 9572−9583

D ow

nl oa

de d

vi a

F L

O R

ID A

I N

T L

U N

IV o

n N

ov em

be r

21 , 2

01 9

at 1

4: 47

:3 8

(U T

C ).

S ee

h tt

ps :/

/p ub

s. ac

s. or

g/ sh

ar in

gg ui

de li

ne s

fo r

op ti

on s

on h

ow t

o le

gi ti

m at

el y

sh ar

e pu

bl is

he d

ar ti

cl es

.

To date, very few studies have investigated AnMBR treatment of only FW,5,6 and no research has evaluated microbial community dynamics within these systems. Notably, there are no prior studies that have applied RNA-based analyses to characterize microbial activity during AnMBR treatment of FW. Given the long SRT in AnMBRs, DNA-based analyses may inaccurately characterize the functional microbial community. It is important to note that comanagement of FW and domestic wastewater via AnMBRs has received attention in the literature recently1,7−9 but is not the focus of this research. Further, few studies have systematically compared performance of SP versus TP10−13 or evaluated the impact of phase separation on the microbial community14,15 during FW treatment via AD (without membrane separation), and there is no prior work investigating TP AnMBR treatment of FW. Although a consensus regarding performance benefits of TP versus SP digestion of FW is lacking,10−13,16 studies have attributed optimized hydrolysis rates and minimization of ammonia accumulation12,16 as key differentiators that resulted in increased methane production in TP. Therefore, in this work we aimed to systematically compare SP and TP management of FW using bench-scale AnMBRs equipped with flat-sheet ceramic membranes. Both systems were operated at incremental OLRs up to 15 g COD L·day−1 to compare performance and operating limits. DNA- and RNA- based high-throughput sequencing were used to evaluate differences in microbial community structure and activity, respectively, between the SP and TP systems across the range of applied OLRs.

2. MATERIALS AND METHODS 2.1. Bench-Scale AnMBR Configuration. Two replicate

jacketed 7 L reactors (Chemglass, NJ) with 5.2 L liquid volume were operated for approximately 300 days (Figure S1). Each system contained a submerged flat-sheet ceramic microfiltration membrane with pore size of 0.1 μm (Cembrane, Denmark) and a total effective membrane area of 0.0113 m2. Both AnMBRs were mixed continuously at 250 rpm with an impellor located near the bottom of the reactor vessel. The jacketed reactors were connected to a recirculating water bath (Fisher Scientific, Hampton, NH) for temperature control to 37 °C with reactor temperature monitored via a probe submerged in the mixed liquor. Pressure in the headspace and permeate lines were monitored using pressure transducers (Transducers Direct, Cincinnati, OH). Influent and permeate were pumped using peristaltic pumps to maintain a constant liquid volume (NewEra, Farmingdale, NY and Langer, Boonton, NJ, respectively). Periodic backwashing (for 2 min every 10 min) and continuous biogas sparging (at a flow rate of 1 m3 m−2·h−1) were employed to manage membrane fouling, where flux ranged from 0.8−1.8 L m−2·h−1 (LMH) for the different feeding rates. For sparging, a minidiaphragm pump (Parker, North Carolina) recirculated produced biogas through sparging tubes transversely mounted below the membrane housing. The headspace was connected to a gas flow meter (GFM) (Aalborg, New York) that continuously measured biogas production. All data acquisition and permeate pump control was done via LabVIEW (National Instruments, Austin, TX), with data recorded in minute intervals. 2.2. Inoculation and Operational Parameters. The

bench-scale AnMBRs were inoculated with mixed liquor collected from a full-scale, mesophilic (37 °C) AnMBR treating FW. Divert, Inc. operates two full-scale AnMBRs for

FW management in the US built by ADI Systems based on Kubota flat-sheet membranes. Their systems are 1.2 and 2.1 M gallons in size and typically operate with OLRs between 1.5− 4.5 g COD L·d−1. The inocula had a total solids (TS) concentration of 28 g·L−1 and a total volatile solids (TVS) concentration of 19 g·L−1. A FW slurry was also collected periodically from Divert, Inc. and filtered using 1 mm aluminum mesh to remove large particles and prevent clogging of tubing in the bench-scale systems. The FW slurry was a blend of processed FW collected from grocery stores in Los Angeles and creamery wastewater, and had an average COD of 122 ± 7 g·L−1. The pH of the feed FW was 3.5 and the total volatile fatty acid (VFA) concentration was 6.3 ± 1.2 g acetic acid equivalent (HAc eq) L−1 (SI Table 2). Initially, both AnMBRs (denoted SP1 and SP2) were operated in SP mode, using replicate operational conditions. After similar perform- ance was observed in both reactors (<10% fluctuations in biogas production and COD removal over greater than 1 month), one AnMBR was transitioned to TP mode by incorporating a TP-AP upstream of the TP-MP AnMBR. The TP-AP was continuously stirred at 250 rpm and partially submerged in a recirculating water bath to maintain a mesophilic temperature of 37 °C. Gas production in the TP- AP was monitored in 10 min intervals using an MPA-200 Methane Potential Analyzer (Challenge Technology, Spring- dale, AR). The HRT and SRT in TP-AP was 3 days throughout the study, which was maintained by increasing the volume of the reactor based on the feeding rate in TP-MP. Next, OLR was incrementally increased in SP and TP-MP from 2.5 to 3.5, 5, 10, and 15 g COD L·d−1. For the lowest OLRs tested, 2.5 and 3.5 g COD L·day−1, FW was diluted to maintain sufficient influent flow rates to prevent operational issues (SI Table 1). In order to achieve the higher OLRs, HRT was decreased in SP and TP-MP. For the 15 g COD L·d−1, a combination of decreased HRT and concentration of influent FW was done to achieve the OLR without significantly increasing membrane flux and associated fouling.

2.3. Chemical Assays and Sampling. Reactor perform- ance was monitored by evaluating influent and effluent characteristics, including COD, VFAs, and ammonia concen- tration. Permeate samples were filtered with 0.2 μm nylon membrane filters (Whatman, Pittsburgh, PA) to measure soluble constituents (ammonium, VFAs, etc.). Total COD analysis was done with high-range dichromate reactor digestion, and pH was measured with a Mettler Toledo probe (Columbus, OH). The Nessler-Method17 was used to determine ammonia concentration. VFAs (formic acid, acetic acid, propionic acid, butyric acid, and valeric acid) and other ions (e.g., nitrate and sulfate) were determined using ion chromatography (ICS-2000, Dionex, Sunnyvale, CA) equipped with a refrigerated autosampler (Thermo Scientific, NY, USA). Chromatographic separation was achieved using a 2 mm AS- 11HC column (Dionex, Sunnyvale, CA). The composition of biogas was measured using the Trace 1310 GC system (Thermo Scientific, NY) equipped with a flame ionization detector (FID) using hydrogen as carrier gas, where a TG- BOND Q 30 m × 0.53 mm × 20 μm column was used for chromatographic separation. TS and TVS of the biomass were determined using procedures outlined in Standard Methods (APHA 2005).

2.5. Microbial Community Analysis. Biomass samples were collected weekly from SP, TP-AP, and TP-MP. In addition to bench-scale AnMBR samples, FW and mixed liquor

Environmental Science & Technology Article

DOI: 10.1021/acs.est.9b02639 Environ. Sci. Technol. 2019, 53, 9572−9583

9573

samples from the full-scale AnMBR were collected monthly over 6 months for DNA-based analyses and weekly over 1 month for RNA-based analyses to understand microbial community structure and activity profiles at the full-scale, which is operated in SP mode. All samples were centrifuged at 5,000g for 5 min at 4 °C, decanted, and preserved at −80 °C until further processing. For RNA preservation, biomass was also stabilized using DNA/RNA shield (Zymo Research, Irvine, CA). Approximately 0.2 g of pelletized biomass was taken for extraction. DNA extraction, RNA extraction, and sequencing were conducted as detailed in our previous work.18

Briefly, DNA extraction was conducted using the Maxwell 16 Blood LEV kit according to manufacturer’s instruction (Promega, Madison, WI), whereas RNA extraction was conducted using the Maxwell 16 simplyRNA blood kit. An additional DNase treatment was conducted using DNA-free DNA Removal Kit (Invitrogen, Carlsbad, CA) to remove DNA contamination from RNA extracts. DNA and RNA quantities were measured using the Quant-iT PicoGreen dsDNA Assay (Invitrogen, Carlsbad, CA) and Quant-iT RiboGreen RNA Assay, respectively. Afterward, reverse transcription to generate single-stranded complementary DNA (cDNA) from RNA extracts was performed using the GoScript Reverse Tran- scription System according to the manufacturer’s instructions (Promega, Madison, WI). One hundred ng of RNA was taken from each sample for cDNA synthesis. Library preparation and sequencing were conducted at the University of Michigan via Illumina MiSeq using the MiSeq Reagent Kit V2 (2 × 250 bp reads) and sequencing primers described previously.19

Sequencing results were analyzed using mothur20 with Silva

13221 as a reference database for alignment and classification. “Relative abundance” is the percentage of 16S rRNA gene sequences (DNA-based) for a given population out of a total of 16S rRNA gene sequences for archaea and bacteria. Likewise, “relative activity” is used to describe the percentage of 16S rRNA sequences (RNA-based) out of a total of 16S rRNA sequences for archaea and bacteria. Statistical analyses were conducted using JMP Pro (SAS Institute, NC) and LEfSe tools.22 All raw sequences form this study are available in NCBI’s Sequence Read Archive (SRA) database23 (BioProject ID PRJNA554898).

3. RESULTS AND DISCUSSION

3.1. TP-AP Effectively Increased VFAs Concentration and Enriched a Distinct Microbial Community from FW. Stable VFA production was achieved in TP-AP, with acetate and propionate constituting the majority of VFAs detected (Figure S2). Initially, propionate concentration was higher than acetate, but acetate later emerged as the dominant VFA after 50 days of operation, which corresponded with the increase of OLR from 2.5 to 3.5 g COD L·day−1. Notably, pH in TP-AP remained relatively constant throughout all OLRs at 3.70 ± 0.39 (Figure S3). We elected to not adjust operating pH to a neutral pH as has been done previously to increase VFA production.24,25 Nonetheless, the total average VFA concentration considering all OLRs in the effluent of the TP- AP was 19.8 ± 6.7 g HAc eq L−1, a substantial increase from the VFA concentration in the FW, 6.3 ± 1.2 g HAc eq L−1

(Table S2).

Figure 1. Methane production rate (primary y-axis) and methane per OLR (secondary y-axis). Solid fill shows SP and pattern fill signifies TP-MP. The number after SP or TP-MP indicates the OLR. For example, TP-MP - 2.5 indicates the mean methane production for TP-MP at 2.5 g COD L· day−1. SP1 and SP2 represent when both AnMBRs were run in SP mode. The solid error bars indicate the 95% confidence interval and the circles indicate standard deviation of daily methane production for each OLR. The two-tailed t test was used to test statistical significance of difference in means between SP and TP-MP for each OLR. The symbol * indicates when mean methane production was significantly different (p < 0.05) between SP and TP-MP for the indicated OLRs.

Environmental Science & Technology Article

DOI: 10.1021/acs.est.9b02639 Environ. Sci. Technol. 2019, 53, 9572−9583

9574

The microbial community analysis revealed distinct community structures and activity profiles between FW and TP-AP (Figure S4). Nonmetric multidimensional scaling (NMDS) and subsequent analysis of molecular variance (AMOVA) were used to determine whether the NMDS clustering was significant, by pooling only FW and TP-AP samples. The difference in microbial communities between FW and TP-MP were statistically significant (p < 0.001), for both RNA- and DNA-based analyses (i.e., 16S rRNA and 16S rRNA genes), as shown from the ANOVA results. Lactobacillus was the most active genus in both FW and TP-AP, representing 71−85% relative activity in FW samples and 64−90% relative activity in TP-AP samples (Figure S5). The dominance of Lactobacillus in TP-AP, a genus known for fermentative metabolism,26 is not surprising given that this population has been shown to withstand low pH conditions. Prominent shifts in the microbial activity profile were

observed with increasing OLR in the TP-AP. Aeriscardovia, a population that was not active in FW (0.23 ± 0.14% relative activity), was highly active at 2.5 g COD L·day−1 with an average of 15.2% relative activity, decreased to <1.30% at 3.5− 10 g COD L·day−1, and increased to 8.38% at 15 g COD L· day−1 (Figure S5). A previous study similarly reported a shift from an Aeriscardovia-dominated community to a Lactobacillus- dominated community at decreased HRT in a TP digester.14

The only known species in the genus, Aeriscardovia aeriphila, have been described as requiring a minimum pH of 4.5 for initial growth.27 Therefore, it is surprising that this genus dominated in our system at a lower pH of 3.70 ± 0.39. Coincidently, as Aeriscardovia became nonactive at 3.5 g COD L·day−1, the TP-AP VFA profile transitioned from being propionate-dominated to acetate-dominated. The low activity of Aeriscardovia at OLRs between 3.5 and 10 g COD L·day−1

corresponded with a substantial increase in activity of acetogenic Acetobacter.28 Although commonly labeled as obligate aerobes,29 Acetobacter have been documented at high relative abundance during FW fermentation to VFAs.30,31

Other populations that were present in high relative activity in FW, such as Pseudomonas and Flavobacterium, continued to show high relative activity in most TP-AP samples. Methanogens and syntrophs were both effectively inhibited in the TP-AP at <0.07% and <0.024% of relative activity, respectively (Figure S6). Low pH and high VFA concen- trations >5.8 g L−1 have been shown to completely inhibit methanogens, eventually also inhibiting the growth of syntrophs that depend on methanogens to maintain low hydrogen partial pressure.32

3.2. TP AnMBR Resulted in Improved Performance Relative to SP AnMBR at High OLRs and >98% COD Removal Efficiency Was Achieved. During the first phase of operation where both AnMBRs were operated in SP mode, methane production was not significantly different between SP1 and SP2 (two-tailed t test, p = 0.71) (Figure 1). COD removal efficiency during this initial phase was >99%, and methane production for SP1 and SP2 was 4.24 ± 0.38 and 4.20 ± 0.36 L d−1, respectively. After this period, SP2 was converted to TP mode (consisting of TP-AP and TP-MP), while SP1 was maintained in SP mode. Significant differences in performance were apparent

between SP and TP-MP as OLR was incrementally increased. At 2.5 g COD L·day−1 both reactors showed similar performance (two-tailed t test, p = 0.85), whereas at 3.5 g COD L·day−1, TP-MP showed significantly higher methane

production (p = 0.00034). Similarly, OLRs of 5 and 10 g COD L·day−1 resulted in significantly higher methane production in TP-MP than SP (Figure 1). At 10 g COD L·day−1, mean methane production in TP-MP was 20.3 ± 8.3% greater (95% confidence interval (CI) of mean) than SP. At each OLR, high variability in daily methane production was apparent for both SP and TP-MP. However, no significant outliers in daily methane production rate were identified (p < 0.05) using Grubbs’ test or the extreme studentized deviate (ESD) method. The highest methane per OLR in SP was observed at 2.5 g COD L·day−1, which was 0.32 ± 0.06 L CH4g COD

−1

fed. The highest specific methane yield per OLR in TP was observed at 3.5 g COD L·day−1, 0.33 ± 0.02 L CH4g COD

−1

fed (Figure 1). The results in our study demonstrated stable performance at higher OLRs and improved methane production in AnMBRs compared to studies that applied conventional ADs for FW treatment, confirming previous reports that systematically compared AnMBRs with ADs.6 A review paper on FW treatment in ADs reported that stable performance was usually attained at OLR lower than 6.4 g COD L·day−1 and HRT of 16−40 days, considering both SP and TP treatment. Further, the highest specific methane reported in our study is higher than studies that applied SP and TP ADs for FW treatment.12,33,34

The COD mass balance indicated that greater COD went to biomass growth and VS accumulation in SP compared to TP- MP (Figure S7), indicating that TP mode improved hydrolysis of FW. At an OLR of 10 g COD L·day−1, the VS concentration in SP reached 43 g L−1 but only 31 g L−1 for TP-MP (Figure S8). The high VS in SP negatively impacted membrane performance, and SRT was subsequently decreased from 140 days (at 2.5 g COD L·day−1) to 32.5 days for SP and 43.3 days for TP-MP to achieve similar VS concentrations in both systems. Methane production per OLR data indicated inhibition in SP at 10 g COD L·day−1, where an abrupt decrease of 26% relative to 2.5 g COD L·day−1 was observed. In TP-MP, only an 11% decrease in methane per OLR was observed. At 15 g COD L·day−1 both systems showed severe inhibition, with 27% and 28% decline in methane per OLR relative to 2.5 g COD L·day−1 (Figure 1). The HRT in the SP and TP-MP ranged between 24 and 10 days for the different OLRs. We did not take into account the additional HRT as a result of the TP-AP reactor in the TP system. We believe this had a relatively minor contribution to performance because of the long total SRT (Table S1). We did not see major differences in effluent COD in SP and TP throughout the different OLRs. In addition, at 15 g COD L·d−1, where the highest difference in HRT occurred due to the additional HRT in TP-AP (1.3 times relative to SP), there was no statistically significant difference in performance between SP and TP. Therefore, we believe the performance differences were more likely linked to SRT rather than HRT. Similarly, a previous study reported that COD removal efficiency in AnMBR during high-strength slaughterhouse waste treatment was independent of HRT and OLR.35

Increased VFA concentration signaled disturbance in SP and TP-MP at high OLRs. At 10 g COD L·day−1, total VFAs increased 33.5-fold in SP and 21.4-fold in TP-MP (Figure S9). Total ammonia nitrogen (TAN) in both systems approached inhibiting concentrations at 10 and 15 g COD L·day−1 (Figure S3). At the highest OLR, TAN reached 1820 mg·L−1 in SP and 1840 mg·L−1 in TP-MP, with calculated free ammonia nitrogen (FAN) concentrations reaching 235 mg·L−1 and 279 mg·L−1 in

Environmental Science & Technology Article

DOI: 10.1021/acs.est.9b02639 Environ. Sci. Technol. 2019, 53, 9572−9583

9575

SP and TP-MP, respectively. Although reported inhibitory ammonia concentrations vary greatly across studies in the literature,36 FAN concentrations as low as 80 mg·L−1 have been shown to cause stress on anaerobic microbial communities.37 It is important to note that a significant decrease in pH was not observed (Figure S3) in either system due to the high alkalinity, 7.8 ± 1.1 and 7.6 ± 1.6 g L−1 as CaCO3 in SP and TP-MP, respectively, at the highest OLR. Effluent COD concentrations increased by 27.9% in SP and 27.8% in TP-MP at 15 g COD L·day−1 compared to 2.5 g COD L·day−1 (Figure S10). However, COD removal efficiency remained high throughout operation, remaining ≥99% for both SP and TP. 3.3. TP Enriched for Syntrophic Fatty-Acid Oxidizing

Bacteria while Stable Methanogen Activity Suggested Functional Redundancy. Higher relative activity of syntrophic fatty-acid oxidizing bacteria was observed in TP- MP compared to SP, whereas methanogens showed similar activity in both systems at all OLRs (Figure 2). We primarily relied on RNA-based sequencing data, due to the higher sensitivity of RNA-based data relative to DNA-based data, particularly when evaluating inhibiting conditions.36 At 10 g COD L·day−1, where the highest difference in performance was

observed between SP and TP-MP, potential syntrophs accounted for 5.1 ± 3.2% relative activity in SP and 10.2 ± 3.3% relative activity in TP-MP. Methanogens represented 19.2 ± 2.1% and 20.5 ± 3.4% relative activity in SP and TP- MP, respectively. These results suggest that methanogens had redundant functionality and were not directly linked to the reduced relative performance of SP. Even at 15 g COD L· day−1, where both systems showed rapid decline in methane production per OLR, methanogens remained at high relative activity, 19.7 ± 2.1% in SP and 26.9 ± 1.8% in TP-MP. Our results are in-line with a review paper38 that theorized based on response of the anaerobic microbiome to disturbances that methanogens could be resistant (endure changes), resilient (rebound after inhibition), and redundant (replace population with similar functionality). However, acetogens/syntrophs were classified as only resistant and resilient and are highly “function-specialized”,38 corroborating our observed positive correlation between syntroph activity and performance. Notably, syntrophs were more abundant/active in our study compared to other studies that applied ADs for FW treatment,39,40 which is significant considering the crucial role syntrophs play in reactor stability by removing VFAs.38

The sustained high activity of methanogens in both SP and

Figure 2. (A) Relative activity of syntrophic fatty-acid oxidizers and (B) relative activity of methanogens identified at the genus level where possible using 16S rRNA sequencing for SP and TP-MP at increasing OLRs. Results are expressed as a percentage normalized using a total of 16S rRNA sequences (Bacteria and Archaea). Truncated y-axes (0 to 12% and 0 to 50% on (A) and (B), respectively) are shown to accentuate differences in activity. The secondary y-axis and pink diamond shaped data points for (A) and (B) signify the ratio of TP-MP to SP for the total sum of syntrophs and methanogens, respectively. The results shown here are average relative activity values for three sampling points for each OLR (only two sampling points for SP1 and SP2).

Environmental Science & Technology Article

DOI: 10.1021/acs.est.9b02639 Environ. Sci. Technol. 2019, 53, 9572−9583

9576

TP-MP indicates that inhibition of other key populations led to the poor performance at 10 g COD L·day−1 in SP and 15 g COD L·day−1 in TP-MP. Methanosaeta were the most active methanogens in both systems, which was also observed in mixed liquor samples from the full-scale AnMBR (Figure S11), where Methanosaeta accounted for 19.5 ± 0.03% relative activity. It is not surprising that Methanosaeta dominated in our study and at the full-scale AnMBR over Methanosarcina, given their higher affinity for acetate41 and relatively low acetate concentrations in the AnMBRs. 3.4. Microbial Activity Data Revealed Increased

Diversity in TP-MP at High OLRs and Distinct Community Profile. In evaluating the microbial community data, we focused on three questions: (i) were there distinct differences in microbial community dynamics between SP and TP-MP; (ii) which microbial communities in SP and TP-MP were resilient at higher OLRs in the presence of potential inhibitors (e.g., ammonia, salt, and VFAs); and (iii) which communities were linked with increased performance at high OLRs. We utilized inverse Simpson index, NMDS analysis, genus-level classification, Linear Discriminant Analysis (LDA), and Spearman Rank analyses to address the above three questions. Inverse Simpson index indicated higher diversity in TP-MP

for all OLRs ≥ 3.5 g COD L·day−1, relative to SP (Figure 3). A

linear decline in diversity with increasing OLR was observed in SP, whereas the TP-MP community had relatively stable diversity throughout operation. NMDS analysis showed some spatial segregation between SP and TP communities (Figure S12). ANOVA analysis of the ordination indicated that the microbial communities in SP and TP-MP were statistically significantly different for the respective DNA- and RNA-based data (p < 0.001). Initially, when both systems were operated in SP mode, they showed similar ordination (SP1 and SP2; Figure S12). However, with increasing OLRs, the difference in SP and TP-MP ordination became apparent. In the TP-AP

community, which had relatively low diversity, the DNA- and RNA-based data clustered together, and ANOVA analysis confirmed that the differences in ordination were not statistically significant (p = 0.19). RNA-based microbial activity data was first screened to only

evaluate genera present in ≥3% relative activity in at least one sample, and the resultant genera subdivided into three distinct groups according to their average relative activity change from 2.5 g COD L·day−1 (SP1/SP2) to an OLR of 10 or 15 g COD L·day−1: (1) decreased by ≥50%, (2) showed ≤50% change, and (3) increased by ≥50% (Figure 4). One of the most notable trends in Group 1 was the increase of Leptotrichiaceae with increasing OLRs, particularly in SP. Leptotrichiaceae relative activity was only 2.27 ± 0.48% during the initial phase (SP1 at 2.5 g COD L·day−1) but increased to 10.6 ± 1.7% at 15 g COD L·day−1(Figure 4). Although the family Leptotrichiaceae, generally known to metabolize a wide range of carbohydrates including disaccharides,42 have been identi- fied in other similar studies,43,44 they are poorly characterized in engineered anaerobic systems. Another important trend from Group 1 communities identified in SP was the substantial increase of Pelolinea, which was not observed in TP-MP. Pelolinea, a recently classified genus from subseafloor sedi- ments,45 accounted for 2.23 ± 0.85% relative activity initially and increased to 4.73 ± 2.10% and 6.40 ± 0.73% at 10 and 15 g COD L·day−1, respectively. Pelolinea are filamentous chemoorganotrophs that ferment sugars.45 To our knowledge, we are the first study to report the proliferation of this genus in engineered anaerobic systems. It is possible that salt accumulation at higher OLRs could have provided a selective pressure resulting in increased activity of Pelolinea as they are known to be halotolerant.45 In our study, chloride concen- trations in SP biomass at 15 g COD L·day−1 OLR exceeded 2 g·L−1. The genera that showed the most substantial increase in

relative activity in TP-MP were Lactobacillus and Treponema (Figure 4). Treponema increased from 0.37 ± 0.06% in SP2 to 3.53 ± 3.02% at 15 g COD L·day−1 in TP-MP. Treponema isolates from termite guts have been characterized as homoacetogens via the Wood-Ljungdahl pathway, while rumen Treponema strains have been described as fermenters able to degrade polysaccharides and disaccharides.46,47

Treponema have been found to dominate during high acetate concentrations,48 and the possibility of this population being syntrophic acetate oxidizers has previously been theorized.49

Also noteworthy, Treponema have been shown to persist in high TAN of up 10 g·L−1,50 and this could have benefited their activity levels in TP-MP. Unclassified bacteria and Proteobac- teria also steeply increased in TP-MP with increasing OLR (each with >7% of the relative activity at 15 g COD L·day−1), indicating that at higher OLR, yet to be described populations likely play an important role in how these systems adapt to inhibition. To characterize microbial populations that were resilient to

inhibition at high OLRs (10 or 15 g COD L·day−1), we used nonparametric analysis with the LEfSe22 tool for all genera ≥0.5% relative activity in at least one sample (Figures S13 and S14). The LEfSe tool applies an LDA analysis, where a score of ≥2 is deemed significant (Figure 5). Our results indicated that three methanogens in TP-MP, Methanoculleus, Methanospir- illum, and Methanomassiliicoccus, showed differential activity at 15 g COD L·day−1. Conversely, Methanoculleus and Meth- anospirillum were more prominent in low OLRs (2.5−3.5 g

Figure 3. Inverse Simpson index of 16S rRNA gene sequencing results for SP, TP-MP, and TP-AP at different OLRs. The index results shown here are average values for three sampling points for each OLR (only two sampling points for SP1 and SP2), and the error bars indicate the standard deviation for the three samples.

Environmental Science & Technology Article

DOI: 10.1021/acs.est.9b02639 Environ. Sci. Technol. 2019, 53, 9572−9583

9577

COD L·day−1) in SP. In TP-MP, potential syntrophs, unclassified Syntrophaceae and Candidatus Cloacimonas, showed high activity at 15 g COD L·day−1. A genomic reconstruction study51 on Candidatus Cloacimonas indicated that this population is a hydrogen producing syntroph. Significant LDA scores at high OLRs in TP-MP were also observed for two populations with limited characterization in engineered anaerobic systems, Pedosphaeraceae and Phaselicys- tis. A single genome study52 reported similar sequences of Pedosphaeraceae and found that phylotypes of Verrucomicrobia within this family significantly contributed to polysaccharide degradation, and we speculate that a similar phenomenon could explain their high activity in TP-MP. For SP, only five genera, Anaerolineaceae, Methanomicro-

biales, Pelolinea, Syntrophaceae, and Treponema, showed significant LDA scores indicating differential grouping at high

OLRs (10 or 15 g COD L·day−1), suggesting that there was more widespread inhibition of active microbial populations in SP (Figure 5). Candidatus Caldatribacterium, Methanobacte- rium, and Methanoculleus were found to be prominent taxa at low OLRs, with the highest LDA scores. In contrast, Methanobacterium and Methanoculleus were among taxa that exhibited differential grouping at high OLRs in TP-MP. Candidatus Caldatribacterium and unclassified Atribacteria JS1 were present at significant LDA at low OLRs in both SP and TP-MP, suggesting that these populations may be sensitive to high OLR regardless of SP or TP mode. Candidatus Caldatribacterium belongs in the recently proposed phylum Atribacteria JS1.53 These populations are linked with possible syntrophic propionate metabolism,54 and their disappearance in TP-MP could be associated with the observed transition from propionate- to acetate-dominated TP-AP (Figure S2).

Figure 4. (A) Relative activity for genera that showed ≥3% relative activity in at least one sample in SP and (B) TP-MP samples. Three distinct groups were formed based on relative activity change in either 10 or 15 g COD L·day−1 relative to SP1/SP2, with increased OLR: (1) decreased by ≥50% (red-fill), (2) showed ≤50% change (blue-fill), and (3) increased by ≥50% (green-fill). Results are expressed as a percentage normalized using a total of 16S rRNA sequences (Bacteria and Archaea). Truncated y-axes (0 to 90%) are shown to accentuate differences in activity. The results shown here are average relative activity values for three sampling points for each OLR (only two sampling points for SP1 and SP2).

Environmental Science & Technology Article

DOI: 10.1021/acs.est.9b02639 Environ. Sci. Technol. 2019, 53, 9572−9583

9578

Although these populations have also been shown to metabolize acetate,54 they may have been outcompeted by other acetate scavengers. Spearman Rank analysis revealed that several of the

populations that correlated with methane production were methanogens or confirmed/suspected acetogens/syntrophs (Figure 6). Populations that were present prominently in both SP and TP-MP and showed significant (p < 0.05) positive correlation with methane production were Spirochaetaceae M2PT2-76 termite group, Blvii28 wastewater-sludge group unclassified Proteobacteria, and Syntrophaceae. The only described species in the genus Blvii28_wastewater-sludg- e_group, Acetobacteroides hydrogenigenes, has been described as a fermenter able to utilize a range of carbohydrates.55,56

Overall, the comparative microbial activity results shed light on which populations were resistant to high OLRs in SP and TP-MP. We were also able to link the improved performance of TP relative to SP for OLRs 3.5−10 g COD L·day−1 to changes in microbial community dynamics, i.e., increased and stable diversity (inverse Simpson index), higher number of taxa that were resistant to increased OLR conditions (LDA scores), and increased activity of specific populations, such as Lactobacillus and Treponema (genera that showed ≥50% increase at high OLRs). Further, it is important to note that several populations, highly correlated with methane production and also displaying significant differential grouping at high

OLRs, have not been well characterized in the literature on anaerobic systems. For example, Spirochaetaceae M2PT2- 76_termite_group have only been described in the termite gut microbiome yet showed significant correlation with methane production in both SP and TP-MP. Other populations with similar trends were Phaselicystis, Pedosphaer- aceae, Paludibacteraceae, and Leptotrichiaceae. This indicates that future research should be dedicated to understanding the underlying competitive advantage these populations may have during inhibiting conditions. Our results are somewhat limited by our reliance on relative activity and abundance data, as opposed to absolute data. In addition, more advanced approaches that link specific substrate uptake or metabolism with detected microbial communities, such as fluorescent reporters and labeled substrates,36 could better elucidate how these systems adjust to operating changes. Future studies should apply more advanced molecular tools to provide a mechanistic understanding of community resistance to operating parameters such as OLR.

3.5. Biofilm Community Showed Higher Activity of Methanogens and Distinct Community Relative to Biomass Samples. Transmembrane pressure (TMP) re- mained low, between 2 and 10 kPa for both SP and TP-MP, during the majority of operation. TMP > 10 kPa was first observed after 224 and 193 days of operation in SP and TP- MP, respectively, with maximum TMP recorded at 48 kPa for

Figure 5. (A) Taxonomic cladogram for populations that showed significant differential relative activity in one of four categories, 2.5−3.5 g COD L·day−1, 5 g COD L·day−1, 10 g COD L·day−1, or 15 g COD L·day−1 OLRs for SP and (B) TP-MP samples. All taxa presented here showed significant differential activity by resulting in LDA scores of ≥2. The analysis was conducted with the LEfSe tool, and relative activity data was used as input for all groups that showed ≥0.5% relative activity in at least one sample. Relative activity results from three sampling points for each OLR were used for the analysis.

Environmental Science & Technology Article

DOI: 10.1021/acs.est.9b02639 Environ. Sci. Technol. 2019, 53, 9572−9583

9579

SP and 45 kPa for TP-MP. Fouling was found to be reversible with chemical cleaning, using hydrochloric acid (0.1 M) and sodium hypochlorite (5% W/V). We observed that there was significantly increased reversible fouling with increased TVS and TS, requiring more frequent chemical cleaning at OLRs > 10 g COD L·day−1, which necessitated the reduction of SRT to maintain operable TS concentrations (Table S1). Six membrane cleaning cycles were conducted throughout the study. Notably, long-term operation and chemical cleaning did not compromise ceramic membrane integrity based on visual observations and effluent quality, which has been an opera- tional challenge in full-scale AnMBRs equipped with polymeric membranes. Biomass samples taken from the biofilm/ membrane foulant layer (10 g COD L·day−1 OLR) indicated similar community structure and activity profiles in SP and TP- MP but a distinct community compared to the respective

biomass samples (Figure S15). NMDS revealed similar ordination for the biofilm DNA- and RNA-based analyses for both SP and TP-MP (Figure S12). Therefore, we grouped these samples together to conduct a nonparametric t test to identify communities that showed significant differential representation. Comparing SP and TP-MP biofilm commun- ities indicated that Geobacter had significantly differential representation in the SP biofilm, whereas Treponema was more present in the TP-MP biofilm. Biofilm methanogenic relative activity was 40% and 56% greater in biofilm samples relative to biomass samples for SP and TP-MP, respectively. This correlates with observations on similar dates where biomass samples showed higher VFA concentrations relative to effluent samples (Figure S9). Thus, the biofilm community aided in improving permeate quality, as also indicated by the high COD removal efficiency, even at high OLRs (Figure S10). A similar

Figure 6. (A) Relative activity of communities that showed significant (p < 0.05) correlation with methane production in SP and (B) TP-MP, identified at the genus level where possible based with 16S rRNA sequencing. Results are expressed as a percentage normalized using a total of 16S rRNA sequences (Bacteria and Archaea). Truncated y-axes (0 to 50%) are shown to accentuate differences in activity. The Spearman correlation coefficient values (ρ) are shown as the x-axis on the legend. The secondary y-axis and blue diamond shaped data points for (A) and (B) signify the mean methane production per day for each OLR. The Spearman analysis was done on triplicate sampling points for each OLR (duplicate for SP1 and SP2 conditions only), where methane production for that specific sampling day and relative activity data were used as inputs. Average relative activity data for the three sampling points per each OLR is shown for simplification (only two sampling points for SP1 and SP2).

Environmental Science & Technology Article

DOI: 10.1021/acs.est.9b02639 Environ. Sci. Technol. 2019, 53, 9572−9583

9580

observation has been reported during AnMBR treatment of low-strength wastewater, where biofilm development signifi- cantly improved permeate quality by reducing effluent acetate and propionate concentrations.57

Prominent communities that showed a significant differ- ential presence in the SP biofilm relative to biomass were Anaerolineaceae and Synergistaceae, with 78% and 342% increase, respectively. Synergistaceae (243% increase) and Syntrophomonas (212% increase) were significantly more present in the TP-MP biofilm relative to biomass. Various studies have reported co-occurrence of Anaerolineaceae along- side methanogens,58,59 prompting discussion of this genus as potential exoelectrogens that coexist with Methanosaeta.58−60

In addition, an omics study61 showed populations within Anaerolineae having substantially active type IV pili, which enable cellular attachment and provide a competitive advantage during attached-growth mode. OTUs within this class had relative activity of 18% and 7.7% in SP and TP-MP biofilm samples, respectively. Although speculative, the overall higher activity of Geobacter and members of Anaerolineae in SP compared to TP-MP in both biofilm and biomass samples suggests that direct interspecies electron transfer (DIET) could be an adaptive mechanism during SP treatment at higher OLRs, whereas TP-MP systems enrich activity of traditional syntrophic populations. We have demonstrated that TP relative to SP mode

improves AnMBR treatment of FW, particularly at high OLRs. Our study demonstrated (1) increased methane production of up to 20.3% at high OLRs, (2) significant enrichment of syntrophic bacteria, (3) increased diversity at high OLRs likely resulting in a more resilient community in the presence of microbial inhibitors, and (4) improved hydrolysis (and VFA production) by enriching for fermentative bacteria in the AP at low pH, while effectively inhibiting methanogens and syntrophs. In addition, we demonstrated that ceramic membranes can be applied during long-term FW treatment without nonreversible fouling. Identifying higher operational limits has important implications because of the expected increase in FW recycling in California and elsewhere as new regulations mandate FW landfill diversion. Higher OLR increases the economic favorability of anaerobic treatment systems, by increasing energy production and minimizing reactor footprint and capital costs. Future studies should compare economic trade-offs in AnMBR FW management, particularly the added costs of incorporating a membrane unit. TP AnMBRs are worth considering to increase efficiency of decentralized anaerobic treatment of FW.

■ ASSOCIATED CONTENT *S Supporting Information The Supporting Information is available free of charge on the ACS Publications website at DOI: 10.1021/acs.est.9b02639.

Additional details on AnMBR operation, performance, and microbial community data (PDF)

■ AUTHOR INFORMATION Corresponding Author *Phone: +1 213.740.0473. E-mail: [email protected]. ORCID Adam L. Smith: 0000-0002-3964-7544 Notes The authors declare no competing financial interest.

■ ACKNOWLEDGMENTS This work was conducted with funding from the National Science Foundation grant no. 1605715. Y.M.A. was partially supported by the Teh Fu Yen Fellowship from the University of Southern California.

■ REFERENCES (1) Becker, A. M.; Yu, K.; Stadler, L. B.; Smith, A. L. Co- management of domestic wastewater and food waste: A life cycle comparison of alternative food waste diversion strategies. Bioresour. Technol. 2017, 223, 131−140. (2) Smith, A.; Shimada, T.; Raskin, L. A comparative evaluation of community structure in full-scale digesters indicates that two-phase digesters exhibit greater microbial diversity than single-phase digesters. Environmental Science: Water Research & Technology 2017, 3 (2), 304−311. (3) Bhattacharya, S. K.; Madura, R. L.; Walling, D. A.; Farrell, J. B. Volatile solids reduction in two-phase and conventional anaerobic sludge digestion. Water Res. 1996, 30 (5), 1041−1048. (4) Bowles, E. C.; Novak, J. T.; Bott, C. B.; Pruden, A. Performance of Acid-Gas Anaerobic Digesters for Minimization of Siloxane and Hydrogen Sulfide in the Biogas. Proceedings of the Water Environment Federation 2012, 2012 (2), 639−653. (5) Xiao, X.; Huang, Z.; Ruan, W.; Yan, L.; Miao, H.; Ren, H.; Zhao, M. Evaluation and characterization during the anaerobic digestion of high-strength kitchen waste slurry via a pilot-scale anaerobic membrane bioreactor. Bioresour. Technol. 2015, 193, 234−242. (6) Cheng, H.; Hiro, Y.; Hojo, T.; Li, Y.-Y. Upgrading methane fermentation of food waste by using a hollow fiber type anaerobic membrane bioreactor. Bioresour. Technol. 2018, 267, 386−394. (7) Zamorano-Loṕez, N.; Moñino, P.; Borraś, L.; Aguado, D.; Barat, R.; Ferrer, J.; Seco, A. Influence of food waste addition over microbial communities in an Anaerobic Membrane Bioreactor plant treating urban wastewater. J. Environ. Manage. 2018, 217, 788−796. (8) Cho, K.; Jeong, Y.; Seo, K. W.; Lee, S.; Smith, A. L.; Shin, S. G.; Cho, S.-K.; Park, C. Effects of changes in temperature on treatment performance and energy recovery at mainstream anaerobic ceramic membrane bioreactor for food waste recycling wastewater treatment. Bioresour. Technol. 2018, 256, 137−144. (9) Jeong, Y.; Hermanowicz, S. W.; Park, C. Treatment of food waste recycling wastewater using anaerobic ceramic membrane bioreactor for biogas production in mainstream treatment process of domestic wastewater. Water Res. 2017, 123, 86−95. (10) Shen, F.; Yuan, H.; Pang, Y.; Chen, S.; Zhu, B.; Zou, D.; Liu, Y.; Ma, J.; Yu, L.; Li, X. Performances of anaerobic co-digestion of fruit & vegetable waste (FVW) and food waste (FW): single-phase vs. two- phase. Bioresour. Technol. 2013, 144, 80−85. (11) Ganesh, R.; Torrijos, M.; Sousbie, P.; Lugardon, A.; Steyer, J. P.; Delgenes, J. P. Single-phase and two-phase anaerobic digestion of fruit and vegetable waste: Comparison of start-up, reactor stability and process performance. Waste Manage. 2014, 34 (5), 875−885. (12) Voelklein, M.; Jacob, A.; O’Shea, R.; Murphy, J. Assessment of increasing loading rate on two-stage digestion of food waste. Bioresour. Technol. 2016, 202, 172−180. (13) Jo, Y.; Kim, J.; Hwang, K.; Lee, C. A comparative study of single-and two-phase anaerobic digestion of food waste under uncontrolled pH conditions. Waste Manage. 2018, 78, 509−520. (14) Shin, S. G.; Han, G.; Lim, J.; Lee, C.; Hwang, S. A comprehensive microbial insight into two-stage anaerobic digestion of food waste-recycling wastewater. Water Res. 2010, 44 (17), 4838− 4849. (15) Wu, Y.; Wang, C.; Liu, X.; Ma, H.; Wu, J.; Zuo, J.; Wang, K. A new method of two-phase anaerobic digestion for fruit and vegetable waste treatment. Bioresour. Technol. 2016, 211, 16−23. (16) Sahu, N.; Deshmukh, S.; Chandrashekhar, B.; Sharma, G.; Kapley, A.; Pandey, R. Optimization of hydrolysis conditions for minimizing ammonia accumulation in two-stage biogas production

Environmental Science & Technology Article

DOI: 10.1021/acs.est.9b02639 Environ. Sci. Technol. 2019, 53, 9572−9583

9581

process using kitchen waste for sustainable process development. J. Environ. Chem. Eng. 2017, 5 (3), 2378−2387. (17) APHA Standard Methods for the Examination of Water and Wastewater, 21st ed.; American Public Health Association: Wash- ington, DC, 2005. (18) Amha, Y. M.; Sinha, P.; Lagman, J.; Gregori, M.; Smith, A. L. Elucidating microbial community adaptation to anaerobic co- digestion of fats, oils, and grease and food waste. Water Res. 2017, 123, 277−289. (19) Kozich, J. J.; Westcott, S. L.; Baxter, N. T.; Highlander, S. K.; Schloss, P. D. Development of a dual-index sequencing strategy and curation pipeline for analyzing amplicon sequence data on the MiSeq Illumina sequencing platform. Appl. Environ. Microbiol. 2013, 79 (17), 5112−5120. (20) Schloss, P. D.; Westcott, S. L.; Ryabin, T.; Hall, J. R.; Hartmann, M.; Hollister, E. B.; Lesniewski, R. A.; Oakley, B. B.; Parks, D. H.; Robinson, C. J. Introducing mothur: open-source, platform- independent, community-supported software for describing and comparing microbial communities. Applied and environmental micro- biology 2009, 75 (23), 7537−7541. (21) Pruesse, E.; Quast, C.; Knittel, K.; Fuchs, B. M.; Ludwig, W.; Peplies, J.; Glöckner, F. O. SILVA: a comprehensive online resource for quality checked and aligned ribosomal RNA sequence data compatible with ARB. Nucleic Acids Res. 2007, 35 (21), 7188−7196. (22) Segata, N.; Izard, J.; Waldron, L.; Gevers, D.; Miropolsky, L.; Garrett, W. S.; Huttenhower, C. Metagenomic biomarker discovery and explanation. Genome biology 2011, 12 (6), R60. (23) Leinonen, R.; Sugawara, H.; Shumway, M. The sequence read archive. Nucleic Acids Res. 2011, 39, D19−D21. (24) Chen, Y.; Jiang, X.; Xiao, K.; Shen, N.; Zeng, R. J.; Zhou, Y. Enhanced volatile fatty acids (VFAs) production in a thermophilic fermenter with stepwise pH increase − Investigation on dissolved organic matter transformation and microbial community shift. Water Res. 2017, 112, 261−268. (25) Khan, M. A.; Ngo, H. H.; Guo, W.; Chang, S. W.; Nguyen, D. D.; Varjani, S.; Liu, Y.; Deng, L.; Cheng, C. Selective production of volatile fatty acids at different pH in an anaerobic membrane bioreactor. Bioresour. Technol. 2019, 283, 120−128. (26) Rault, A.; Bouix, M.; Beál, C. Fermentation pH influences the physiological-state dynamics of Lactobacillus bulgaricus CFL1 during pH-controlled culture. Appl. Environ. Microbiol. 2009, 75 (13), 4374− 4381. (27) Simpson, P. J.; Ross, R. P.; Fitzgerald, G. F.; Stanton, C. Bifidobacterium psychraerophilum sp. nov. and Aeriscardovia aeriphila gen. nov., sp. nov., isolated from a porcine caecum. Int. J. Syst. Evol. Microbiol. 2004, 54 (2), 401−406. (28) Iizuka, H.; Komagata, K. Taxonomy of genus Pseudomonas with special reference to their modes of metabolism of carbon compounds. J. Gen. Appl. Microbiol. 1963, 9 (1), 83−95. (29) Fukaya, M.; Park, Y.; Toda, K. Improvement of acetic acid fermentation by molecular breeding and process development. J. Appl. Bacteriol. 1992, 73 (6), 447−454. (30) Cao, J.; Zhang, Q.; Wu, S.; Luo, J.; Wu, Y.; Zhang, L.; Feng, Q.; Fang, F.; Xue, Z. Enhancing the anaerobic bioconversion of complex organics in food wastes for volatile fatty acids production by zero- valent iron and persulfate stimulation. Sci. Total Environ. 2019, 669, 540−546. (31) Stra ̈uber, H.; Bühligen, F.; Kleinsteuber, S.; Dittrich- Zechendorf, M. Carboxylic acid production from ensiled crops in anaerobic solid-state fermentation-trace elements as pH controlling agents support microbial chain elongation with lactic acid. Engineering in Life Sciences 2018, 18 (7), 447−458. (32) Xu, Z.; Zhao, M.; Miao, H.; Huang, Z.; Gao, S.; Ruan, W. In situ volatile fatty acids influence biogas generation from kitchen wastes by anaerobic digestion. Bioresour. Technol. 2014, 163, 186− 192. (33) Voelklein, M. A.; O’Shea, R.; Jacob, A.; Murphy, J. D. Role of trace elements in single and two-stage digestion of food waste at high organic loading rates. Energy 2017, 121, 185−192.

(34) Grimberg, S.; Hilderbrandt, D.; Kinnunen, M.; Rogers, S. Anaerobic digestion of food waste through the operation of a mesophilic two-phase pilot scale digester−assessment of variable loadings on system performance. Bioresour. Technol. 2015, 178, 226− 229. (35) Jensen, P.; Yap, S.; Boyle-Gotla, A.; Janoschka, J.; Carney, C.; Pidou, M.; Batstone, D. Anaerobic membrane bioreactors enable high rate treatment of slaughterhouse wastewater. Biochem. Eng. J. 2015, 97, 132−141. (36) Amha, Y. M.; Anwar, M. Z.; Brower, A.; Jacobsen, C. S.; Stadler, L. B.; Webster, T. M.; Smith, A. L. Inhibition of anaerobic digestion processes: applications of molecular tools. Bioresour. Technol. 2018, 247, 999−1014. (37) Werner, J. J.; Garcia, M. L.; Perkins, S. D.; Yarasheski, K. E.; Smith, S. R.; Muegge, B. D.; Stadermann, F. J.; DeRito, C. M.; Floss, C.; Madsen, E. L. Microbial community dynamics and stability during an ammonia-induced shift to syntrophic acetate oxidation. Appl. Environ. Microbiol. 2014, 80 (11), 3375−3383. (38) Carballa, M.; Regueiro, L.; Lema, J. M. Microbial management of anaerobic digestion: exploiting the microbiome-functionality nexus. Curr. Opin. Biotechnol. 2015, 33, 103−111. (39) Li, L.; He, Q.; Ma, Y.; Wang, X.; Peng, X. A mesophilic anaerobic digester for treating food waste: process stability and microbial community analysis using pyrosequencing. Microb. Cell Fact. 2016, 15 (1), 65. (40) Amha, Y. M.; Sinha, P.; Lagman, J.; Gregori, M.; Smith, A. L. Elucidating microbial community adaptation to anaerobic co- digestion of fats, oils, and grease and food waste. Water Res. 2017, 123, 277−289. (41) Conklin, A.; Stensel, H. D.; Ferguson, J. Growth kinetics and competition between Methanosarcina and Methanosaeta in meso- philic anaerobic digestion. Water Environ. Res. 2006, 78 (5), 486−496. (42) Lory, S. The family Leptotrichiaceae. Prokaryotes: Firmicutes and Tenericutes 2014, 213−214. (43) Ziels, R. M.; Svensson, B. H.; Sundberg, C.; Larsson, M.; Karlsson, A.; Yekta, S. S. Microbial rRNA gene expression and co- occurrence profiles associate with biokinetics and elemental composition in full-scale anaerobic digesters. Microb. Biotechnol. 2018, 11 (4), 694−709. (44) Stolze, Y.; Bremges, A.; Rumming, M.; Henke, C.; Maus, I.; Pühler, A.; Sczyrba, A.; Schlüter, A. Identification and genome reconstruction of abundant distinct taxa in microbiomes from one thermophilic and three mesophilic production-scale biogas plants. Biotechnol. Biofuels 2016, 9 (1), 156. (45) Imachi, H.; Sakai, S.; Lipp, J. S.; Miyazaki, M.; Saito, Y.; Yamanaka, Y.; Hinrichs, K.-U.; Inagaki, F.; Takai, K. Pelolinea submarina gen. nov., sp. nov., an anaerobic, filamentous bacterium of the phylum Chloroflexi isolated from subseafloor sediment. Int. J. Syst. Evol. Microbiol. 2014, 64 (3), 812−818. (46) Bekele, A. Z.; Kobayashi, Y.; Koike, S. Phylogenetic diversity and dietary association of rumen Treponema revealed using group- specific 16S rRNA gene-based analysis. FEMS Microbiol. Lett. 2011, 316 (1), 51−60. (47) Paster, B. J.; Canale-Parola, E. Treponema saccharophilum sp. nov., a large pectinolytic spirochete from the bovine rumen. Appl. Environ. Microbiol. 1985, 50 (2), 212−219. (48) Xie, X.; Yang, C.; Guan, L. L.; Wang, J.; Xue, M.; Liu, J. X. Persistence of Cellulolytic Bacteria Fibrobacter and Treponema After Short-Term Corn Stover-Based Dietary Intervention Reveals the Potential to Improve Rumen Fibrolytic Function. Front. Microbiol. 2018, 9, 1363. (49) Ahlert, S.; Zimmermann, R.; Ebling, J.; König, H. Analysis of propionate-degrading consortia from agricultural biogas plants. MicrobiologyOpen 2016, 5 (6), 1027−1037. (50) Poirier, S.; Desmond-Le Queḿeńer, E.; Madigou, C.; Bouchez, T.; Chapleur, O. Anaerobic digestion of biowaste under extreme ammonia concentration: identification of key microbial phylotypes. Bioresour. Technol. 2016, 207, 92−101.

Environmental Science & Technology Article

DOI: 10.1021/acs.est.9b02639 Environ. Sci. Technol. 2019, 53, 9572−9583

9582

(51) Pelletier, E.; Kreimeyer, A.; Bocs, S.; Rouy, Z.; Gyapay, G.; Chouari, R.; Rivier̀e, D.; Ganesan, A.; Daegelen, P.; Sghir, A.; Cohen, G. N.; Med́igue, C.; Weissenbach, J.; Le Paslier, D. Candidatus Cloacamonas acidaminovorans”: genome sequence reconstruction provides a first glimpse of a new bacterial division. Journal of bacteriology 2008, 190 (7), 2572−2579. (52) Martinez-Garcia, M.; Brazel, D. M.; Swan, B. K.; Arnosti, C.; Chain, P. S.; Reitenga, K. G.; Xie, G.; Poulton, N. J.; Gomez, M. L.; Masland, D. E. Capturing single cell genomes of active polysaccharide degraders: an unexpected contribution of Verrucomicrobia. PLoS One 2012, 7 (4), No. e35314. (53) Dodsworth, J. A.; Blainey, P. C.; Murugapiran, S. K.; Swingley, W. D.; Ross, C. A.; Tringe, S. G.; Chain, P. S.; Scholz, M. B.; Lo, C.- C.; Raymond, J. Single-cell and metagenomic analyses indicate a fermentative and saccharolytic lifestyle for members of the OP9 lineage. Nat. Commun. 2013, 4, 1854. (54) Nobu, M. K.; Dodsworth, J. A.; Murugapiran, S. K.; Rinke, C.; Gies, E. A.; Webster, G.; Schwientek, P.; Kille, P.; Parkes, R. J.; Sass, H. Phylogeny and physiology of candidate phylum ‘Atribacter- ia’(OP9/JS1) inferred from cultivation-independent genomics. ISME J. 2016, 10 (2), 273. (55) Su, X.-L.; Tian, Q.; Zhang, J.; Yuan, X.-Z.; Shi, X.-S.; Guo, R.- B.; Qiu, Y.-L. Acetobacteroides hydrogenigenes gen. nov., sp nov., an anaerobic hydrogen-producing bacterium in the family Rikenellaceae isolated from a reed swamp. Int. J. Syst. Evol. Microbiol. 2014, 64 (1), 2986−2991. (56) Su, X.-L.; Tian, Q.; Zhang, J.; Yuan, X.-Z.; Shi, X.-S.; Guo, R.- B.; Qiu, Y.-L. Acetobacteroides hydrogenigenes gen. nov., sp nov., an anaerobic hydrogen-producing bacterium in the family Rikenellaceae isolated from a reed swamp. Int. J. Syst. Evol. Microbiol. 2014, 64, 2986. (57) Smith, A. L.; Skerlos, S. J.; Raskin, L. Membrane biofilm development improves COD removal in anaerobic membrane bioreactor wastewater treatment. Microb. Biotechnol. 2015, 8 (5), 883−894. (58) McIlroy, S. J.; Kirkegaard, R. H.; Dueholm, M. S.; Fernando, E.; Karst, S. M.; Albertsen, M.; Nielsen, P. H. Culture-Independent Analyses Reveal Novel Anaerolineaceae as Abundant Primary Fermenters in Anaerobic Digesters Treating Waste Activated Sludge. Front. Microbiol. 2017, 8, 1134. (59) Xu, H.; Wang, C.; Yan, K.; Wu, J.; Zuo, J.; Wang, K. Anaerobic granule-based biofilms formation reduces propionate accumulation under high H2 partial pressure using conductive carbon felt particles. Bioresour. Technol. 2016, 216, 677−683. (60) Wang, G.; Li, Q.; Gao, X.; Wang, X. C. Synergetic promotion of syntrophic methane production from anaerobic digestion of complex organic wastes by biochar: Performance and associated mechanisms. Bioresour. Technol. 2018, 250, 812−820. (61) Xia, Y.; Wang, Y.; Wang, Y.; Chin, F. Y. L.; Zhang, T. Cellular adhesiveness and cellulolytic capacity in Anaerolineae revealed by omics-based genome interpretation. Biotechnol. Biofuels 2016, 9 (1), 111.

Environmental Science & Technology Article

DOI: 10.1021/acs.est.9b02639 Environ. Sci. Technol. 2019, 53, 9572−9583

9583