1 Pyrosequencing analysis yields comprehensive assessment of microbial communities 2 in pilot-scale two-stage Membrane Biofilm Reactors 3 4 Aura Ontiveros-Valencia1,2, Youneng Tang1,3, He-Ping Zhao1,4, David Friese5, Ryan 5 Overstreet5, Jennifer Smith6, Patrick Evans6, Bruce E. Rittmann1, Rosa Krajmalnik- 6 Brown1*. 7 8 1 9 University, 1001 South McAllister Ave. Tempe, AZ 85287-5701 USA Biodesign Institute, Swette Center for Environmental Biotechnology, Arizona State 10 2 School of Sustainability, Arizona State University 11 3 University of Illinois at Urbana-Champaign, Urbana, IL 61801, USA (current affiliation) 12 4 13 Environmental and Resource Science, Zhejiang University, Hangzhou, China (current 14 affiliation) 15 5 APTwater, Inc., 100 W. Broadway, Suite 200, Long Beach, California 90802 16 6 CDM-Smith, 14432 SE Eastgate Way, Bellevue, Washington 98007 MOE Key Lab of Environmental Remediation and Ecosystem Health, College of 17 18 19 *Corresponding author contact information: Telephone 1-480-727-7574, dr.rosy@asu.edu 20 21 22 23 Email: dr.rosy@asu.edu 24 Abstract 25 We studied the microbial community structure of pilot two-stage Membrane Biofilm 26 Reactors (MBfRs) designed to reduce nitrate (NO3-) and perchlorate (ClO4-) in 27 contaminated groundwater. The groundwater also contained oxygen (O2) and sulfate 28 (SO42-), which became important electron sinks that affected the NO3- and ClO4- removal 29 rates. Using pyrosequencing, we elucidated how important phylotypes of each “primary” 30 microbial group –denitrifying bacteria (DB), perchlorate-reducing bacteria (PRB), and 31 sulfate-reducing bacteria (SRB) -- responded to changes in electron-acceptor loading. 32 UniFrac, principal coordinate analysis (PCoA), and diversity analyses documented that 33 the microbial community of biofilms sampled when the MBfRs had a high acceptor 34 loading were phylogenetically distant from and less diverse than the microbial 35 community of biofilm samples with lower acceptor loadings. Diminished acceptor 36 loading led to SO42- reduction in the lag MBfR, and this allowed Desulfovibrionales (an 37 SRB) and Thiothrichales (sulfur-oxidizers) to thrive through S cycling. Due to this 38 cooperative relationship, they competed effectively with DB/PRB phylotypes such as 39 Xanthomonadales and Rhodobacterales. Thus, pyrosequencing illustrated that, while 40 DB, PRB, and SRB responded predictably to changes in acceptor loading, a decrease in 41 total acceptor loading led to important shifts within the “primary” groups, the onset of 42 other members (e.g. Thiothrichales), and overall greater diversity. 43 44 Keywords: pilot MBfR, nitrate, perchlorate, sulfate, pyrosequencing (deep sequencing), 45 community structure, community function. 46 Introduction Nitrate (NO3-) is a prevalent water contaminant due to its heavy use in fertilizers 47 48 and widespread presence in wastewater. NO3- can cause methemoglobinemia1,2 in infants 49 and spur eutrophication in water bodies. NO3- is regulated by the US EPA,3 which 50 established a maximum contaminant level (MCL) of 10 mg N/L for drinking water. 51 Perchlorate (ClO4-) is an oxyanion with great chemical stability and is a constituent of 52 rocket propellants, fireworks, and explosives. ClO4-, a normally recalcitrant contaminant 53 found in waters of 35 US states and Puerto Rico,4 can disrupt the thyroid after ingestion. 54 Although ClO4- is not yet listed as a regulated chemical,5 the USEPA is planning to issue 55 an MCL.6 NO3- and ClO4- often are found together at contaminated sites, because 56 ammonium nitrate (NH4NO3), ammonium perchlorate (NH4ClO4), and potassium nitrate 57 (KNO3) are used together for the production of rocket fuel and explosives.4 Destruction of NO3- and ClO4- by microbial respiration has been well 58 59 documented.7-10 NO3- reduction can enhance or hinder ClO4- reduction11-14 depending on 60 the operating conditions of bioremediation approaches. Particularly, the inhibition of 61 ClO4- reduction originates from the competition between denitrifying bacteria (DB) and 62 perchlorate-reducing bacteria (PRB) for common resources, such as the electron donor,15 63 space in biofilms,15 and reductase enzymes.16-18 However and regardless of possible 64 complications, simultaneous microbial respiration of NO3- and ClO4- has been reported.19- 65 20 66 even more pressing when in addition to NO3- and ClO4-, other electron acceptors such as 67 sulfate (SO42-) also are present in the water to be treated. Furthermore, the need to manage the microbial communities in the system becomes 68 3 69 The presence of SO42- in a NO3-- and ClO4- -contaminated groundwater was the 70 situation encountered during demonstration of a pilot two-stage Membrane Biofilm 71 Reactor (MBfR) system.21 In the MBfR, hydrogen gas (H2) diffuses through the walls of 72 hollow-fiber membranes and is used as electron donor by microorganisms that grow as a 73 biofilm on the membranes while utilizing oxidized compounds present in the water 74 flowing throw the reactor as electron acceptors.22 Previous research with MBfR biofilms 75 pointed out competitive relationships between NO3- and ClO4- reductions for which a 76 NO3- loading above 0.6 g N/m2 day at a fixed H2-delivery capacity slowed ClO4- 77 reduction.15 78 Based on the desire to minimize competition between NO3- and ClO4- 79 reductions11-18 when the groundwater to be remediated had a high NO3- : ClO4- ratio (~76 80 g N: 1 g ClO4-), Evans et al.21 set up a two-stage pilot-scale MBfR. The lead MBfR 81 treated the raw groundwater and performed the bulk of denitrification. This lowered the 82 NO3- loading and the potential for NO3- reduction to compete with ClO4- reduction in the 83 lag MBfR, which received the effluent from the lead MBfR.21 The strategy was mostly 84 successful, since most of the NO3- removal occurred in the lead MBfR; however, the two- 85 stage pilot MBfR could not consistently drive the ClO4- concentrations to below the 86 detection limit of 4 µg/L.21 87 In an initial effort to understand the pilot MBfR’s performance, Zhao et al.23 88 assessed the microbial community structure of the pilot reactors using the quantitative 89 Polymerase Chain Reaction (qPCR) targeting characteristic reductases. DB (determined 90 by the nitrite reductases nirK and nirS) were the most abundant microbial group; 91 however, sulfate-reducing bacteria (SRB) (quantified by the dissimilatory sulfite 4 92 reductase dsrA) became dominant and may have outnumbered DB in the pilot MBfRs 93 when the NO3- + O2 loading was low, below 0.3 g H2/m2 day.23 PRB (quantified by the 94 perchlorate-reductase pcrA) were the smallest microbial fraction and were adversely 95 affected when SRB became important, a finding consistent with previous bench-scale 96 studies.24 97 In contrast to these pilot results, Ontiveros-Valencia et al.25 was able to achieve 98 complete ClO4- reduction in a two-stage bench-scale MBfR, even though the ClO4- 99 concentration was unusually high (~4000 µg/L) and SO42- was amply present (~55-60 100 mg/L). The success was attributed to an effective management of the microbial ecology 101 of the reactors so that SO42- reduction was minimized, especially in the lag MBfR. 102 Ontiveros-Valencia et al.25 suppressed SRB in the lag MBfR by re-oxygenating the 103 influent to the lag MBfR to increase the total-acceptor loading and by lowering the H2 104 availability by either decreasing the H2 pressure or by using a less-H2 permeable 105 membrane. Neither strategy was followed with the pilot two-stage MBfR system: Re- 106 oxygenation of the effluent from the lead MBfRs was not possible with the pilot 107 configuration, and the pilot-MBfRs were mostly run with excess H2 availability to 108 encourage ClO4- reduction.21 109 Added to the fact that treatment is more challenging when SO42- is present in the 110 water to be treated, only limited information is available on the ecological interactions 111 between SRB and PRB. Waller26 suggested that the microbial community structure of 112 consortia explored in her study was responsible for the decline in ClO4- reduction when 113 high SO42- concentration was available. However, other studies reported no effect from 5 114 SO42- on ClO4- microbial reduction27-28. Thus, more research addressing these critical 115 ecologic interactions is needed. 116 Although Zhao et al.23 provided a broad view of the “primary” respiratory groups 117 (i.e., DB, PRB, and SRB) in the pilot MBfRs, we employ high-throughput 118 pyrosequencing to gain a deeper understanding of the microbial community structure, 119 including more insight into the phylotypes that constitute the primary respiratory groups 120 present when NO3-, ClO4-, and SO42- are the electron acceptors and a view of other 121 members within the biofilm. 122 Our study addresses the ecological interactions among DB, PRB, SRB, and other 123 microbial groups that developed during bioremediation of groundwater polluted with 124 NO3- and ClO4- with SO42- also present. In particular, we use UniFrac and principal 125 coordinate analysis (PCoA)29,30 to demonstrate that distinctly different communities 126 developed in the biofilm when the acceptor-loading rate was decreased significantly. 127 Furthermore, we explore how decreased acceptor loading led to shifts within the primary 128 members and the development of important other members (e.g., heterotrophs and sulfur- 129 oxidizing bacteria) in the community. While Zhao et al.23 used qPCR to provide an 130 analysis of community structure according to the primary respiratory groups, our findings 131 discriminate among conditions significantly altering the community structure, making the 132 biofilm more diverse and causing shifts within and outside the primary microbial groups. 133 6 134 Materials and Methods 135 MBfR configuration and performance 136 Detailed information about the pilot-MBfRs configuration is given by Evans et 137 al.21 and Zhao et al.23 In brief, the two-stage MBfR was composed of two 500-gallon 138 (1890-L) vessels containing 4 MBfR modules with membrane surface area of 144 m2 per 139 module. The manufacture and on-site configuration of the pilot-MBfR modules was 140 done by APTwater and CDM-Smith. Figure 1a shows that the pilot-MBfR modules were 141 cylindrical and made of woven fabric of polypropylene fibers, which formed sheets of 142 fibers wrapped around a perforated acrylonitrile butadiene styrene (ABS) core. Each 143 module contained ~140,000 polypropylene fibers (200µm OD, Teijin, LTD, Japan). H2 144 gas diffused through the fiber sheet, and water passed through the perforations in the 145 ABS core. The lead and lag MBfRs also were equipped with a set of side reactors for 146 taking biofilm samples without disturbing the biofilm in the modules.21,23 Figure 1b&c 147 shows the side reactors with their connections for water and H2. 148 The pilots were set up to treat a site historically used for munitions and explosives 149 manufacture and surroundings agricultural fields. Hence, the oxidized contaminants in 150 the groundwater were NO3- at 8-9 mg N/L and ClO42- at 160-200 µg/L. The influent also 151 contained O2 at ~8 mg/L and SO42- at ~22 mg/L. The lead and lag positions were 152 switched every 3 days to make the biofilm development similar in both MBfRs and with 153 the goal of minimizing the abundance of SRB in the lag MBfR.21 The H2 pressure and 154 influent flow rate were adjusted according to the conditions in Table 1. The four 155 conditions are representative periods of continuous operation of the pilot system. 156 Adjustment of the influent flow rate led to a proportional change in the total electron- 7 157 acceptor surface loading: Conditions 3 and 4 had significantly lower total electron 158 acceptor loadings than did Conditions 1 and 2. The use of an excess H2-delivery capacity 159 was done to ensure good NO3- removal in the lead MBfR and to achieve complete ClO4- 160 reduction in the lag MBfR.21 161 Samples were collected for off-site analysis at Test America (Irvine, CA), which 162 is certified by the California Environmental Laboratory Accreditation Program (ELAP). 163 The off-site assessment involved measurements for the lead and lag concentrations of 164 NO3- and SO42- (US EPA method 300) and ClO4- (US EPA 314); they were performed 165 three, one, and three times per week, respectively. In addition, measurements for NO3- 166 and sulfide (as a surrogate for SO42- reduction) were carried out three times per week on- 167 site using field kits (CHEMetrics, Virginia, USA).21 O2 and pH were measured by a hand 168 held probes.21 The pH during operation was maintained between 7.4-7.8. The maximum 169 H2 delivery capacity was calculated according to Tang et al. 31 and reported in Table 1. 170 Our work is complementary to the work reported by Zhao et al.23, and both studies are 171 built on the field demonstration described by Evans et al.21 172 Biofilm microbial ecology by pyrosequencing analysis 173 Side reactors representing conditions 1, 2, 3, and 4 were taken after 60, 116, 221, 174 and 263 days of continuous operation, respectively, and were sent in ice containers to the 175 Swette Center for Environmental Biotechnology for microbial community analysis. The 176 samples arrived within 24 hours and were processed according to Zhao et al.23 for DNA 177 extraction. DNA samples were stored at -80°C until shipping for 454 pyrosequencing. 178 DNA samples for 454 pyrosequencing were sent to the Molecular Research DNA lab 179 (Austin, Texas, USA), which performed amplicon pyrosequencing using a standard 8 180 Roche 454/GS-FLX Titanium.32 The Bacteria domain was targeted by selecting the V6 181 and V7 regions of the 16S rRNA gene with primers 939F (5'- 182 TTGACGGGGGCCCGCAC-3') and 1492R (5'TACCTTGTTACGACTT-3').33 We 183 processed the raw data using QIIME 1.7.0 suite34 and removed sequences having fewer 184 than 250 bps, homopolymers of more than 6 bps, primer mismatches, or an average 185 quality score lower than 25. We picked the operational taxonomic unit (OTUs) using the 186 Greengenes 16S rDNA database with uclust 35 based on ≥ 97% identity, removed OTUs 187 that contain less than two sequences (singletons) from our analysis, and aligned the 188 representative sequence of each OTU to the Greengenes Database using PyNast.36,37 189 Potentially chimeric sequences were identified by using ChimeraSlayer,38 and a python 190 script in QIIME was employed to remove the chimeric sequences. We assigned 191 taxonomy to OTUs with BLAST using the SILVA database39 and constructed Newick- 192 formatted phylogenetic trees using FasTree.40 193 For the purpose of eliminating heterogeneity related to having different numbers 194 of sequences among the samples, we sub-sampled the OTU table by randomly selecting 195 ten different times the lowest number of sequences (6800) found among the samples. We 196 then generated PCoA plots and Unweighted Pair Group Method Arithmetic Mean 197 (UPGMA) plots30 using jack-knifed beta diversity. 198 We estimated the OTU richness by calculating Chao1,41 which determines the 199 asymptote on an accumulative curve, predicting how many OTUs would be present if a 200 high number of sequences had been collected, and the phylogenetic relationships by 201 using phylogenetic diversity (PD),42 which estimates the cumulative branch lengths from 202 random OTUs. To evaluate the microbial species diversity and evenness, we computed 9 203 the Shannon43 and Simpson44 indexes. A higher value for the Shannon index indicates 204 greater microbial diversity, while a value for the Simpson metric near one shows an even 205 distribution of bacterial groups within the sample. Sequence data sets are available at 206 NCBI/Sequence Read Archive (SRA) under study with accession number SRP038958. 207 10 208 Results and Discussion 209 Microbial community function 210 Table 2 synthesizes the performance of the pilot-scale reactors. The lead MBfRs 211 were responsible for ~99% of the O2 respiration, 70-90% denitrification, and a small loss 212 of ClO4-.21,23 In the lead MBfRs, the NO3- + O2 flux was greater than ~ 0.34 g H2/m2- 213 day23 (Table 2), which completely suppressed SO42- reduction and is consistent with the 214 bench-scale results of Ontiveros-Valencia et al.45 and modeling work by Tang et al.46 215 Therefore, NO3- and SO42- were the dominant electron acceptors entering the lag MBfR, 216 and the total acceptor surface loading to the lag MBfR was much lower than for the lead 217 MBfR (Table 1). Although the objective of reducing the flow rate and total acceptor 218 loading for Conditions 3 and 4 was to enhance ClO4- removal in the lag MBfR, its major 219 impact was to favor SO42- reduction, an undesired outcome that led to lower ClO4- 220 removal fluxes in the lag MBfR (Table 2). 221 222 223 Electron-acceptor loading affects microbial diversity and structure Table S1 shows all the values for the diversity and evenness metrics for the four 224 conditions. Overall, Chao1, Shannon, and PD values show that the microbial diversity of 225 biofilm samples from Conditions 3 and 4, which had a low acceptor loading (Table 1), 226 was greater than from Conditions 1 and 2, which had a higher acceptor loading. 227 Consistent with the Chao1 results and based on the Simpson index, biofilm samples from 228 Conditions 3 and 4 were more evenly distributed than those in Conditions 1 and 2. 229 230 Figure 2 shows the unweighted UniFrac analysis of the biofilm samples, which is based on the presence or absence of all the phylotypes within a sample. The biofilm 11 231 samples with high acceptor loading (Conditions 1 and 2) clearly formed a cluster (blue 232 branch) distinct from the cluster of Conditions 3 and 4 (red branch). Thus, the large 233 changes in acceptor loading between Conditions 2 and 3 led to very different microbial 234 communities. Particularly for Conditions 1 and 2, the lead and lag biofilms were not 235 significantly different due to the regular switching of positions.21 236 Figure 3 presents the unweighted PCoA plot, which reinforces the clustering 237 found with the UniFrac analysis. The biofilm communities of Conditions 1 and 2 were 238 close to each other along the PC1 vector, while those biofilm samples of Conditions 3 239 and 4 were distant. In an attempt to differentiate the driving force for the PC1 vector, we 240 connect the removal fluxes for SO42- and ClO4- (Table 2) with the community analysis by 241 PCoA. Conditions 3 and 4 had importantly decreased average acceptor loadings (Table 242 1), and SO42- reduction increased significantly (Table 2). The PC1 vector correlates with 243 increased SO42- reduction, particularly from Condition 2 to Condition 3. Hence, the 244 microbial community structure was substantially modified when SO42- reduction became 245 a more important electron sink, a trend also noted by Ontiveros-Valencia et al.33 246 Condition 2 was different from Conditions 1, 3, and 4 along the PC2 vector. This trend is 247 most likely explained by the substantially higher ClO4- flux for Condition 2, which is 248 illustrated in Table 2. 249 While the low electron acceptor loadings primarily shaped the microbial 250 community, particularly by favoring SO42- reduction, operation time also allowed 251 biomass buildup33,45 that may have contributed to structural changes in the biofilm 252 communities. However, operational conditions, such as to the flow rate and hydraulic 253 retention time (HRT), are directly connected to the electron acceptor loadings: 12 254 Decreased flow rate and the consequent higher HRT cause a lowered electron acceptor 255 loading. Extra H2 delivery capacity also can frame the community on its own; however, 256 the excess capacity to deliver electron donor rates was similar across conditions, while 257 the loading of electron acceptor was significantly modified. 258 259 260 Taxonomic breakdown and shifts in the microbial community structure Figure 4 synthesizes the taxonomical break down at the order level of the most 261 abundant phylotypes. Figure S1 also reports the ten most abundant phylotypes for all 262 conditions at the genus level. Consistent with UniFrac and PCoA, the biofilm 263 communities of the lead and lag MBfR were similar for each Condition. The brackets in 264 the legend of Fig. 4 identify the known DB, PRB, SRB, and other types. The groupings 265 show four important trends. First, ~86% of the taxonomic breakdown had microbial 266 phylotypes most closely related to characterized DB and PRB for Condition 1, but these 267 primary groups decreased for subsequent conditions, being only ~60% by Condition 4. 268 Connecting this community trend to community function, DB and PRB phylotypes 269 (reported by pyrosequencing in Figure 4) follow the same trend as the NO3-, O2, and 270 ClO4- fluxes (Table 2). 271 Second, the decrease of microbial phylotypes most closely related to DB and PRB 272 was accompanied by significant increases in microbial phylotypes most closely related to 273 SRB: from <1% in Condition 1 to ~13% in Condition 4. The SRB trend by 274 pyrosequencing is similar to the SRB trend noted by Zhao et al.23 using qPCR; however, 275 the qPCR study found that SRB had become the largest primary group in Condition 4, 276 followed by DB and PRB. It is possible that qPCR overestimated SRB, because some 13 277 DB harbor the dsrA gene.47 Regardless of the method employed, the key trend is that 278 SRB became important with lower acceptor loading. As noted by Ontiveros-Valencia et 279 al.,24 SRB become detrimental to PRB when they are able to occupy the most favorable 280 zones in the biofilm (near the H2-delivering substratum).46 Therefore, incomplete ClO4- 281 reduction in the lag MBfR can be at least partially attributed to increased competition 282 from SRB. 283 Third, lowered electron acceptor loadings leading to augmented SO42- reduction 284 (Conditions 3 and 4) boosted the sulfur-oxidizing Thiotrichales and the SRB 285 Desulfovibrionales. This combination points towards a cooperative relationship based on 286 active S cycling in which Thiotrichales oxidizes H2S produced by SRB while respiring 287 NO3- to ammonia (NH4+). Sulfide oxidation by Thiotrichales provided additional SO42- 288 for SRBs, probably allowing SRB to grow to higher proportions than what would be 289 predicted from the one-time reduction of SO42-. Figure S1 shows that closely related 290 Thiothix phylotypes, which belong to the Thiotrichales order, were abundant at 291 Conditions 3 and 4, and they might have imposed a risk for fouling the membranes due to 292 its filamentous growth.49 Thiothrix can accumulate S granules in its interior from the 293 oxidation of H2S and form rosettes, which are arrangements of filaments. 50-51 Staff 294 operating the pilot MBfRs reported observing filaments in some biofilms. Sulfide 295 oxidizers also were reported in MBfR biofilms by Zhao et al.,52 who observed abundant 296 Campylobacteriales (sulfur-oxidizing bacteria), and by Ontiveros-Valencia et al.,25 who 297 reported significant presence of Ignavibacteriales (green sulfur-oxidizing bacteria) and 298 Thiobacteriales (sulfur-oxidizing bacteria) when SO42- reduction was favored in bench- 299 scale MBfRs. The differences in the phylotypes of the sulfur-oxidizers observed in the 14 300 bench- versus pilot-scale MBfRs probably can be attributed to the different inocula in 301 each study. Despite the different inocula, the cooperative relationship between SRB and 302 sulfur-oxidizing bacteria seems to be common once SO42- reduction becomes important 303 and seems to have accentuated an ecological advantage for SRB. 304 Besides sulfur-oxidizers, heterotrophic microorganisms such as Bacteroidales and 305 Flavobacteriales increased in Conditions 3 and 4. The heterotrophs likely consumed 306 soluble microbial products, whose rate of release increased with high rates of SO42- 307 reduction.33,45 . Likewise, the relative abundance of “unclassified” bacteria and minor 308 phylotypes (microbial groups at <1% abundance) (not shown in Figure 3) went from an 309 average ~3% in Condition 1 to ~8% in Condition 4. The upswing of heterotrophs, 310 unclassified bacteria, and minor phylotypes was the foundation for the increase in the 311 microbial diversity with decreased acceptor loading (Table S1). The greater abundance 312 of other groups and SRB certainly imposed more competition for space in the biofilm, 313 forcing PRB to less favorable positions in the biofilm (zones more likely to detach).24,46 314 Recently, Martin et al.53 employed modeling to explain how increased detachment 315 hindered MBfR performance. Thus, increasing diversity in the biofilm was correlated 316 with poorer performance for ClO4- reduction. 317 Fourth, the DB and PRB groups showed important shifts with acceptor loading. 318 In Conditions 1 and 2, Rhodobacterales were dominant; however, the most abundant DB 319 and PRB phylotypes shifted to Xanthomonadales and Rhodocyclales in Conditions 3 and 320 4. In particular, closely related Aquimonas phylotypes, which belong to the 321 Xanthomonadales order, were common to all biofilm samples, remaining in the biofilm 322 regardless of competition (Fig. S1). In contrast, Rhodobacterales declined dramatically 15 323 in Conditions 3 and 4. Species Rhodobacter capsulatus and Rhodobacter sphaeroides 324 can reduce chlorate (ClO3-) to chlorite (ClO2-); however, no growth was associated with 325 this metabolism.54 326 Other substantial shifts in the phylotypes most closely related to DB and PRB 327 were observable. While the DB and PRB phylotype Rhizobiales remained relatively 328 constant across conditions, the phylotype Hydrogenophilales increased in Conditions 3 329 and 4. Lastly, phylotype Burkholderiales decreased abruptly while phylotype 330 Pseudomonadales decreased slightly. These substantial shifts in the DB and PRB 331 support that the biofilm communities were functionally redundant, which allowed 332 different phylotypes to gain or lose prominence as acceptor loading changed without 333 affecting denitrification performance. 334 In conclusion, pyrosequencing allowed us to comprehensively assess the 335 microbial community diversity and structure of pilot MBfRs. UniFrac, and PCoA helped 336 us understand the main drivers for the shifts in microbial structures. Biofilm 337 communities developed with low total acceptor loading were more diverse and 338 phylogenetic distant from communities with a higher acceptor loading. Primary members 339 (i.e., DB, PRB, and SRB) overall tracked the reduction of the electron acceptors, but 340 showed important shifts with acceptor loading. The DB/PRB phylotype Rhodobacterales 341 was significantly abundant at high acceptor loading; however, the phylotype 342 Xanthomonadales was overall the most dominant DB/PRB phylotype in all biofilm 343 samples. Desulfovibrionales and Thiothrichales appeared together at low acceptor 344 loadings and when SO42- reduction was strong, suggesting S cycling that corresponded to 345 a slowing of the ClO4--reduction rate. Likewise, heterotrophic bacteria became more 16 346 important with lower acceptor loading. The abundance of SRB and sulfur-oxidizing 347 partners, as well as heterotrophs, likely accentuated competition for space and forced 348 PRB to less favorable positions in the biofilm. Thus, the increase in diversity with low 349 acceptor loading was due to the increases in SRB, sulfur-oxidizers, and heterotrophs, and 350 it correlated with poorer performance in terms of ClO4- reduction. 351 352 Acknowledgements This research was funded by the Environmental Security Technology 353 354 Certification Program (ESTCP) by grant ER-200541. The authors express recognition to 355 the Consejo Nacional de Ciencia y Tecnologia (CONACYT) for providing a scholarship 356 to Aura Ontiveros-Valencia towards pursuing her graduate studies and conducting this 357 study. 358 Supporting Information 359 Alpha-diversity metrics and ten most abundant genera in biofilm samples across 360 the four conditions in the pilot system. This information is available free of charge via the 361 Internet at http://pubs.acs.org/ 17 Table 1 Four Conditions identified H2 availability (controlled by H2 pressure) and electron-acceptor surface loadings (adjusted by influent flow rate) for lead and lag MBfRs Condition Flow Hydraulic H2 NO3--N O2 surface SO42ClO4- surface Total Average rate Retention pressure surface loading surface loading electron electron 3 m /d Time loading loading acceptor acceptor surface loading loading atm g H2/m2-d g H2/m2-d g H2/m2-d g H2/m2-d g H2/m2 g H2/m2 day day hours lead lag lead lag lead lag lead lag lead lag lead lag 1 65 0.7 2.2 1.8 0.41 0.13 0.15 0.002 0.22 0.22 0.002 0.002 0.78 0.36 0.6 2 98 0.5 2.8 2.3 0.66 0.17 0.23 0.006 0.33 0.33 0.003 0.002 1.22 0.51 0.9 3 44 1.0 2.2 0.002 0.18 0.18 0.002 0.0004 0.65 0.22 0.4 4 33 1.4 2.1 1.6 0.23 0.02 0.08 0.0004 0.11 0.11 0.001 0.0002 0.41 0.13 0.3 2 0.37 0.03 0.10 We calculated the electron acceptor loading rates according to: 𝑄 × (𝑆 ° ) 𝐿𝑜𝑎𝑑𝑖𝑛𝑔 = (𝑒𝑞. 1) 𝐴 2 where Q = volumetric flow rate (L/day), A = membrane surface area (m ), and S° is the influent concentration (g/L) for an electron acceptor. Each electron acceptor loading value was normalized to g H2/m2 day based on stoichiometric relationships described elsewhere.15-23-25 Total electron-acceptor loading was calculated as the sum of the loadings for O2, NO3-, ClO4-, and SO42-. The average electron acceptor loading was calculated from the lead and lag total electron acceptor loadings at each condition. The lead and lag positions were switched every three days; therefore, an average estimate of the acceptor loading is valuable. The HRT was the same for each reactor regardless of the position. 18 Table 2 Electron acceptor and donor fluxes for lead and lag MBfRs for the four conditions tested over time Condition Nitrate 2 g H2/m day Oxygen flux 2 g H2/m day Sulfate flux Perchlorate flux g H2/m2 day 2 g H2/m day Total H2 experimental flux g H2/m2 day Maximum H2 flux Oversupply of H2 g H2/m2 day g H2/m2 day Lead Lag Lead Lag Lead Lag Lead Lag Lead Lag Lead Lag Lead Lag 1 0.28 0.13 0.15 0.002 0 0.0006 0 0.0008 0.43 0.13 0.57 0.46 0.14 0.3 2 0.49 0.17 0.21 0.004 0 0.001 0.001 0.0018 0.7 0.2 0.72 0.59 0.02 0.4 3 0.24 0.03 0.09 0.002 0 0.0026 0.0007 0.00038 0.33 0.03 0.57 0.51 0.24 0.48 4 0.2 0.02 0.07 0.0004 0 0.003 0.0007 0.00019 0.27 0.02 0.53 0.41 0.26 0.39 The electron acceptor fluxes were reported elsewhere.23 The maximum H2 flux was calculated as Tang et al.31 and the oversupply of H2 corresponded to the maximum H2 flux minus the total H2 experimental flux. 19 Figure 1 a Pilot MBfR module which shows the ABS core and woven fabric. The water and H2 flows are pointed by arrows. b&c show side reactors which were sent to ASU for community analysis. The side reactors were operated as the pilot MBfRs. b shows the water lines feeding the side reactors, and c visualizes the gas connections for the H2 fed, and a closer look of the biofilm in the fiber sheet. 20 Figure 2 Clustering based on the unweighted UniFrac analyses. The branch length represents the distance between biofilm samples in UniFrac units, as indicated by the scale bar. The labels on each branch indicate the biofilm sample of either lead or lag MBfR at the four conditions applied to the reactors. The blue branch correspond to the reactors operated at high electron acceptor surface loadings (Conditions 1 and 2), while the red branch reflect the microbial community performing under low total electron acceptor surface loading (Conditions 3 and 4). 21 Figure 3 Principal Coordinate Analysis (PCoA) based on the unweighted UniFrac. 22 Figure 4 Microbial community structure in lead and lag MBfRs at the order level. The sum does not add up to 100% in all cases because phylotypes < 1% are not shown. The brackets in the legend group the orders according to known members of the noted metabolic groups. DB/PRB phylotypes are shown which hatched fills that clearly show a decline from Condition 1 to Condition 4. Some members of the “heterotrophic microorganisms,” are capable of denitrification under specific circumnstances, such as when using acetate as electron donor and carbon source.48 23 24 References (1) (2) (3) (4) (5) (6) (7) (8) (9) (10) (11) (12) (13) (14) (15) (16) (17) Knobeloch, L.; Salna, B.; Hogan, A.; Postle, J.; Anderson, H. Blue babies and nitrate-contaminated well water. Environmental Health Perspectives, 2000, 108 (7): 675. Camargo, J. A.; Alonso, Á. Ecological and toxicological effects of inorganic nitrogen pollution in aquatic ecosystems: a global assessment. Environment international, 2006, 32(6): 831-849. US EPA. Basic Information about Nitrate in Drinking Water. http://water.epa.gov/drink/contaminants/basicinformation/nitrate.cfm (accessed February 17, 2014). US EPA. Perchlorate treatment technology update. No.EPA 542-R-05-015. 2005. US EPA. Perchlorate. http://water.epa.gov/drink/contaminants/unregulated/perchlorate.cfm (accessed February 17, 2014). US EPA. Potential SBAR Panel: Drinking water regulatory actions for perchlorate. http://www.epa.gov/rfa/perchlorate.html (accessed January 30, 2014). Kim, K.; Logan, B. E. Fixed-Bed bioreactor treating perchlorate-contaminated waters. Environ. Eng. Sci. 2000, 17 (5): 257-265. Lee, K. C.; Rittmann, B. E. Applying a novel autohydrogenotrophic hollow-fiber membrane biofilm reactor for denitrification of drinking water. Water Res. 2002, 36: 2040-2052. Zhang, H.; Bruns, M. A.; Logan, B. E. Perchlorate reduction by a novel chemolithoautotrophic, hydrogen-oxidizing bacterium. Environ. Microbiol. 2002, 4 (10): 570–576. Hatzinger, P. B. Perchlorate biodegradation for water treatment. Environ. Sci. Technol. 2005, 39 (11): 239A-247A. Herman, D. C.; Frankenberger, W. T. 1998. Microbial-mediated reduction of perchlorate in groundwater. J Environ Qual, 1998, 27(4): 750-754. Herman, D. C.; Frankenberger, W. T. Bacterial reduction of perchlorate and nitrate in water. J Environ Qual, 1999, 28(3): 1018-1024. Chaudhuri, S. K.; O'Connor, S. M.; Gustavson, R. L.; Achenbach, L. A.; Coates, J. D. Environmental factors that control microbial perchlorate reduction. Appl Environ Microbiol, 2002, 68(9): 4425-4430. Coates, J. D.; Achenbach, L. A. Microbial perchlorate reduction: rocket fuelled metabolism. Nature Rev Microbiol 2004, 2(7): 569-580. Tang, Y.; Zhao, H.-P.; Marcus, A. K.; Krajmalnik-Brown, R.; Rittmann, B. E. A steady-state biofilm model for simultaneous reduction of nitrate and perchlorate -Part 2: Parameter optimization and results and discussion. Environ Sci Technol 2012, 46(3): 1608-1615. Romanenko, V. I.; Koren'kov, V. N.; Kuznetsov, S. I. Bacterial decomposition of ammonium perchlorate. Mikrobiologiia, 1976, 45(2): 204. Hochstein, L. I.; Tomlinson, G. A. The enzymes associated with denitrification. Annu. Rev. Microbiol. 1988, 42: 231– 261 25 (18) (19) (20) (21) (22) (23) (24) (25) (26) (27) (28) (29) (30) (31) (32) Kengen, S. W. M.; Rikken, G. B.; Hagen, W. R.; van Ginkel, C. G.; Stams, A. J. M. Purification and characterization of (per)-chlorate reductase from the chloraterespiring strain GR-1 J. Bacteriol. 1999, 181: 6706– 6711 Logan, B. E.; Wu, J.; Unz, R. F. Biological perchlorate reduction in high-salinity solutions. Water Res, 2001, 35(12): 3034-3038. Min, B.; Evans, P. J.; Chu, A. K.; Logan, B. E. Perchlorate removal in sand and plastic media bioreactors. Water Res, 2004, 38(1): 47-60. Evans, P.; Smith, J.; Singh, T.; Hyung, H.; Arucan, C.; Berokoff, D.; Friese, D.; Overstreet, R.; Vigo, R.; Rittmann, B. E.; Ontiveros-Valencia, A.; Zhao, H.-P.; Tang, Y.; Kim B.-O., van Ginkel, S.; Krajmalnik-Brown, R. Final Report: Nitrate and Perchlorate Destruction and Potable Water Production Using Membrane Biofilm Reduction. ESTCP Project ER-200541 2013. Rittmann, B. E. The membrane biofilm reactor is a versatile platform for water and wastewater treatment. Environ. Eng. Res. 2007, 12 (4): 157-175. Zhao, H.-P.; Ontiveros-Valencia, A.; Tang, Y.; Kim, B.-O.; van Ginkel, S.; Friese, D.; Overstreet, R.; Smith, J.; Evans, P.; Krajmalnik-Brown, R.; Rittmann, B. E. Removal of multiple electron acceptors by pilot-scale, two-stage membrane biofilm reactors. Water Res. 2014, 54: 115-122. Ontiveros-Valencia, A.; Tang, Y.; Krajmalnik-Brown, R.; Rittmann, B. E. Perchlorate reduction from a highly contaminated groundwater in the presence of sulfate-reducing bacteria in a hydrogen-fed biofilm. Biotechnol Bioeng 2013, 110(12): 3139-3147. Ontiveros-Valencia, A.; Tang, Y; Krajmalnik-Brown, R.; Rittmann, B. E. Managing the interactions between sulfate- and perchlorate-reducing bacteria when using hydrogen-fed biofilms to treat a groundwater with a high perchlorate concentration. Water Res. 2014, 55: 215-224. Waller, S. Bioremediation of perchlorate-contaminated groundwater. Master thesis. University of Toronto. 2002. Attaway, H.; Smith, M. Reduction of perchlorate by an anaerobic enrichment culture. J Ind Microbiol 1993, 12(6): 408-412. Bardiya, N.; Bae, J. H. Bioremediation potential of a perchlorate-enriched sewage sludge consortium. Chemosphere 2005, 58(1): 83-90. Lozupone, C.; Knight, R. UniFrac: a new phylogenetic method for comparing microbial communities. App. Environ. Microbiol. 2005, 71: 8228-8235. Lozupone, C.; Hamady, M.; Knight, R. UniFrac - an online tool for comparing microbial community diversity in a phylogenetic context. BMC Bioinformatics 2006, 7(1): 371. Tang, Y.; Zhou, C.; Van Ginkel, S.; Ontiveros-Valencia, A.; Shin, J.; Rittmann, B. E. Hydrogen-Permeation Coefficients of the Fibers Used in H2-Based Membrane Biofilm Reactors. J. Membrane Sci. 2012, 407-408: 176-183. Sun, Y.; Wolcott, R. D.; Dowd, S. E. Tag-encoded FLX amplicon pyrosequencing for the elucidation of microbial and functional gene diversity in any environment. High-Throughput Next Generation Sequencing. Meth. Mol. Biol. 2011, 733: 129-141. 26 (33) (34) (35) (36) (37) (38) (39) (40) (41) (42) (43) (44) (45) (46) Ontiveros-Valencia, A.; Ilhan, Z. E.; Kang, D.-W.; Rittmann, B. E.; KrajmalnikBrown, R. Phylogenetic analysis of nitrate- and sulfate-reducing bacteria in a hydrogen-fed biofilm. FEMS Microbiol. Ecol. 2013, 85: 158-167. Caporaso, J. G.; Kuczynski, J.; Stombaugh, J.; Bittinger, K.; Bushman, F. D.; Costello, E. K.; Fierer, N.; Pena, A. G.; Goodrich, J. K.; Gordon, J. I.; Huttley, G. A.; Kelley, S. T.; Knight, D.; Koenig, J. E.; Ley, R. E.; Lozupone, C. A.; McDonald, D.; Muegge, B. D.; Pirrung, M.; Reeder, J.; Sevinsky, J. R.; Tumbaugh, P. J.; Walters, W. A.; Widmann, J.; Yatsunenko, T.; Zaneveld, J.; Knight, R. Qiime allows analysis of high-throughput community sequencing data. Nat. Methods 2010, 7: 335-336. Edgar, R. C. Search and clustering orders of magnitude faster than blast. Bioinformatics 2010, 26: 2460-2461. DeSantis, T. Z.; Hugenholtz, P.; Larsen, N.; Rojas, M.; Brodie, E. L.; Keller, K.; Huber, T.; Dalevi, D.; Hu, P.; Andersen, G. L. Greengenes, a chimera-checked 16s rrna gene database and workbench compatible with arb. Appl. Environ. Microbiol. 2006, 72: 5069-5072. Caporaso, J. G.; Bittinger, K.; Bushman, F. D.; DeSantis, T. Z.; Andersen, G. L.; Knight, R. Pynast: A flexible tool for aligning sequences to a template alignment. Bioinformatics 2010, 26: 266-267. Haas, B. J.; Gevers, D.; Earl, A. M.; Feldgarden, M.; Ward, D. V.; Giannoukos, G.; Ciulla, D.; Tabaa, D.; Highlander, S. K.; Sordergren, E.; Methé, B.; DeSantis, T. Z.; The Human Microbiome Consortium, Petrosino, J. F.; Knight, R.; Birren, B. W. Chimeric 16s rrna sequence formation and detection in sanger and 454pyrosequenced pcr amplicons. Genome Res. 2011, 21: 494-504. Pruesse, E.; Quast, C.; Knittel, K.; Fuchs, B. M.; Ludwig, W.; Peplies, J.; Gloeckner, F. O. Silva. A comprehensive online resource for quality checked and aligned ribosomal RNA sequence data compatible with ARB. Nucleic Acids Res. 2007, 35: 7188-7196. Price, M. N.; Dehal, P. S.; Arkin, A. P. FastTree: computing large minimum evolution trees with profiles instead of a distance matrix. Mol. Biol. Evol. 2009, 26: 1641-1650. Hughes, J. B.; Hellmann, J. J.; Ricketts, T. H.; Bohannan, B. J. Counting the uncountable: statistical approaches to estimating microbial diversity. Appl. Environ. Microbiol. 2001, 67 (10): 4399-4406. Faith, D. P. Conservation evaluation and phylogenetic diversity. Biol. Conserv. 1992, 61: 1-10. Shannon, C. E. A mathematical theory of communication. Bell System Technical Journal. 1948, 27: 379–423. Simpson, E. H. Measurement of diversity. Nature 1949, 163: 688. Ontiveros-Valencia, A.; Ziv-El, M.; Zhao, H.-P.; Feng, L.; Rittmann, B. E.; Krajmalnik-Brown, R. Interactions between nitrate-reducing and sulfate-reducing bacteria coexisting in a hydrogen-fed biofilm. Environ. Sci. Technol. 2012, 46: 11289-11298. Tang, Y.; Ontiveros-Valencia, A.; Liang, F.; Zhou, C.; Krajmalnik-Brown, R.; Rittmann, B. E. A biofilm model to understand the onset of sulfate reduction in denitrifying membrane biofilm reactors. Biotechnol. Bioeng. 2012, 110: 763-772. 27 (47) (48) (49) (50) (51) (52) (53) (54) Wu, W.-M.; Gu, B.; Fields, M. W.; Gentile, M.; Ku, Y.-K.; Yan, H.; Tiquias, S.; Yan, T.; Nyman, J.; Zhou, J.; Jardine, P. M.; Craig, C. S. Uranium reduction by denitrifying biomass. J. Bioremed. 2005, 9 (1): 41-61. Adav, S. S.; Lee, D. J.; Lai, J. Y. Microbial community of acetate utilizing denitrifiers in aerobic granules. Applied microbiology and biotechnology, 2010, 85(3), 753-762. Madigan, M.; Markinko, J.; Stahl, D.; Clark, D. Brock Biology of microorganisms; 12th ed.; Pearson: San Francisco California 2009. Williams, T. M.; Unz, R. F. Isolation and characterization of filamentous bacteria present in bulking activated sludge. Appl. Microbiol. Biotechnol. 1985, 22 (4): 273282. Williams, T. M.; Unz, R. F.; Doman, J. T. Ultrastructure of Thiothrix spp. and “type 021N” bacteria. Appl. Environ. Microbiol. 1987, 53 (7): 1560-1570. Zhao, H.-P.; Ilhan, Z. E.; Ontiveros-Valencia, A.; Tang, Y.; Rittmann, B. E.; Krajmalnik-Brown, R. Effects of multiple electron acceptors on microbial interactions in a hydrogen-based biofilm. Environ. Sci. Technol. 2013, 47: 73967403. Martin, K. J.; Picioreanu, C.; Nerenberg, R. Multidimensional modeling of biofilm development and fluid dynamics in a hydrogen-based, membrane biofilm reactor (MBfR). Water Res. 2013, 47 (13): 4739-4751. Roldan, M. D. Chlorate and nitrate reduction in phototrophic bacteria Rhodobacter capsulatus and Rhodobacter sphaeroides. Curr. Microbiol. 1994, 29: 241-245 28