Higher Substrate Ratios of Ethanol to Acetate Steered Chain

Oct 18, 2018 - Thus, care should be taken not to overload the system with primarily ethanol as the substrate and to lower the organic loading rate...
0 downloads 0 Views 2MB Size
Subscriber access provided by RMIT University Library

Sustainability Engineering and Green Chemistry

Higher Substrate Ratios of Ethanol to Acetate Steered Chain Elongation Toward n-Caprylate in a Bioreactor with Product Extraction Catherine Spirito, Alex Marzilli, and Largus T. Angenent Environ. Sci. Technol., Just Accepted Manuscript • DOI: 10.1021/acs.est.8b03856 • Publication Date (Web): 18 Oct 2018 Downloaded from http://pubs.acs.org on October 21, 2018

Just Accepted “Just Accepted” manuscripts have been peer-reviewed and accepted for publication. They are posted online prior to technical editing, formatting for publication and author proofing. The American Chemical Society provides “Just Accepted” as a service to the research community to expedite the dissemination of scientific material as soon as possible after acceptance. “Just Accepted” manuscripts appear in full in PDF format accompanied by an HTML abstract. “Just Accepted” manuscripts have been fully peer reviewed, but should not be considered the official version of record. They are citable by the Digital Object Identifier (DOI®). “Just Accepted” is an optional service offered to authors. Therefore, the “Just Accepted” Web site may not include all articles that will be published in the journal. After a manuscript is technically edited and formatted, it will be removed from the “Just Accepted” Web site and published as an ASAP article. Note that technical editing may introduce minor changes to the manuscript text and/or graphics which could affect content, and all legal disclaimers and ethical guidelines that apply to the journal pertain. ACS cannot be held responsible for errors or consequences arising from the use of information contained in these “Just Accepted” manuscripts.

is published by the American Chemical Society. 1155 Sixteenth Street N.W., Washington, DC 20036 Published by American Chemical Society. Copyright © American Chemical Society. However, no copyright claim is made to original U.S. Government works, or works produced by employees of any Commonwealth realm Crown government in the course of their duties.

Page 1 of 32

Environmental Science & Technology

1

Higher Substrate Ratios of Ethanol to Acetate Steered Chain Elongation Toward n-

2

Caprylate in a Bioreactor with Product Extraction

3 4

Catherine M. Spirito1, Alexander M. Marzilli1, Largus T. Angenent1,2*

5 6

1

7

14853, USA

8

2

Department of Biological and Environmental Engineering, Cornell University, Ithaca, NY

Center for Applied Geosciences, University of Tübingen, 72074 Tübingen, Germany

9 10

*Correspondence to:

11

Largus T. Angenent

12

Email: [email protected]

13

1 ACS Paragon Plus Environment

Environmental Science & Technology

14

ABSTRACT

15

Syngas fermentation to ethanol and acetate has recently been coupled to microbial chain

16

elongation to produce medium-chain carboxylates (MCCs), including n-caproate and n-

17

caprylate. These medium chain carboxylates are relatively hydrophobic, and thus easier to

18

extract from solution than miscible ethanol. Here, we examined the effect of eleven different

19

ethanol-to-acetate substrate ratios (ranging from 1.8 to 275 g COD g COD-1 [1.2 to 183 mole

20

mole-1]) on directing chain elongation toward n-caprylate in a 0.7-L upflow anaerobic filter with

21

product extraction. During an eight-month operating period, we monitored the performance and

22

characterized the microbiome composition of this chain-elongating bioreactor. We also

23

developed a thermodynamic model to predict the favorability of n-caprylate production at

24

different substrate ratios. As predicted by our model, higher ethanol-to-acetate substrate ratios

25

fed to our bioreactor led to higher specificities for n-caprylate production. We observed that

26

feeding primarily ethanol to the bioreactor (i.e., ethanol-to-acetate substrate ratio of 275 g COD

27

g COD-1) resulted in the highest specificity for n-caprylate, but the n-caprylate production rate

28

decreased at this high ratio, resulting in lower conversion efficiencies. Thus, care should be taken

29

not to overload the system with primarily ethanol as the substrate and to lower the organic

30

loading rate.

31 32

INTRODUCTION

33

For syngas fermentation, pyrolysis gas (a mixture of CO, CO2, and H2 gas) from the gasification

34

of organic waste materials (e.g., municipal solid waste or industrial waste), as well as off-gas

35

from steel manufacturing, can be converted into ethanol by anaerobic syngas-fermenting bacteria

36

(i.e., acetogens).1-4 Ethanol production rates of higher than 0.14 mole L-1 h-1 (6 g L-1 h-1) have

2 ACS Paragon Plus Environment

Page 2 of 32

Page 3 of 32

Environmental Science & Technology

37

been reported for syngas fermentation systems.2,

5

38

commercial syngas fermentation plant went on-line in China in the spring of 2018,6 and the

39

company is currently constructing a plant in Belgium.7 Besides ethanol, acetate is another 2-

40

carbon compound (C2) that is present in the effluent of syngas fermentation systems and the

41

relative product ratio for ethanol and acetate is tunable depending on the operating conditions.2

42

For optimized syngas fermentation systems, the ethanol-to-acetate product ratio can be high,

43

with reported ratios that are typically ~30 (based on chemical oxygen demand [COD] with the

44

unit g COD g COD-1; or a molar ratio of 20 mole mole-1).8, 9 In a recent study by Abubackar et

45

al.,10 the production of primarily ethanol via syngas fermentation by Clostridium

46

autoethanogenum DSM 10061 was reported. Importantly, the concentration of ethanol produced

47

by syngas fermentation is typically limited to 1.3 M (6% w/v) due to the product toxicity of

48

ethanol to acetogens.2

The company Lanzatech announced that a

49 50

Extraction of this dilute ethanol from syngas fermentation broth via distillation is energetically

51

costly. A study by Vane et al.11 estimated that for fractional distillation of 1.3-M ethanol

52

solutions, 0.23 MJ-fuel is required per mole of ethanol distilled. To circumvent energy-intensive

53

ethanol distillation, an alternative route was found by chain elongating ethanol into medium-

54

chain carboxylates (MCCs), such as n-caproate (C6) and n-caprylate (C8).12 This has been

55

performed by pure-, co-, or open-cultures by placing syngas fermentation and chain elongation in

56

series,13-15 or by combining all pathways into one-stage bioreactor.16-18 With open cultures of

57

microbial consortia (reactor microbiomes),19-21 the pH of the broth can be lowered to mildly

58

acidic conditions (5-5.5)12,

59

conditions the MCCs exist at relatively high concentrations as undissociated carboxylic acid

15

to inhibit acetoclastic methanogens. At these mildly acidic

3 ACS Paragon Plus Environment

Environmental Science & Technology

Page 4 of 32

60

species (pKa ~ 4.9),22 which have a low maximum solubility concentration. In-line product

61

extraction systems can be employed to maximize chain elongation by removing the accumulating

62

undissociated medium-chain carboxylic acids, which can become toxic to bacteria.23 The

63

coupling of a mildly acidic broth and an in-line product extraction system should result in the

64

extraction of the product from the broth at much lower energy consumption rates than ethanol

65

distillation, though an in-depth energy consumption analysis has not been conducted. To avoid

66

addition of acid or base to the bioreactor and extraction system, an in-line membrane electrolysis

67

system (i.e., an electrochemical system) was employed by Xu et al.24 following the extraction

68

unit of the bioreactor. By using renewable electric power, n-caproic acid can be separated from

69

the extraction solution by taking advantage of the innate pH gradient between the anode and

70

cathode of the membrane electrolysis system.

71 72

An electron donor, such as ethanol or lactic acid, is required to provide the energy and carbon to

73

chain elongate by the reverse β-oxidation pathway. Syngas fermentation broth, as mentioned

74

previously, has a relatively high ethanol-to-acetate substrate ratio. Previous pure-culture studies

75

of reverse β-oxidation with Clostridium kluyveri have shown that higher ethanol-to-acetate

76

substrate ratios led to higher n-caproate-to-n-butyrate product ratios in both batch25,

77

continuously-fed27 bioreactors. Open-culture studies have also found that higher ethanol

78

concentrations can drive chain elongation toward n-caproate, as long as the ethanol concentration

79

does not become inhibitory.28-30 A recent study by Kucek et al.15 demonstrated via batch tests

80

that increasing concentrations of ethanol compared to acetate also led to increased production of

81

n-caprylate compared to n-caproate. A simple generalized stoichiometric model demonstrated

82

that increased ethanol-to-acetate substrate ratios created more thermodynamically favorable

4 ACS Paragon Plus Environment

26

and

Page 5 of 32

Environmental Science & Technology

83

conditions for chain elongation to n-caproate,20 which explained the higher n-caproate-to-n-

84

butyrate product ratios that others had observed (see above). This previous model by Angenent et

85

al.20 did not include n-caprylate production.

86 87

Here, our main objective was to test eleven different ethanol-to-acetate substrate ratios ranging

88

from 1.8 to 275 g COD g COD-1 (1.2 to 183 mole mole-1) in a continuously operated chain-

89

elongating bioreactor with product extraction to steer the production toward n-caprylate. In

90

addition, we investigated whether n-caprylate could be produced from primarily ethanol in the

91

substrate (i.e., an ethanol-to-acetate substrate ratio 275 g COD g COD-1), which had not

92

previously been demonstrated. We extended the thermodynamic model by Angenent et al.20 to n-

93

caprylate and we used our experimental data to validate the model. Finally, we characterized the

94

bioreactor microbiome via Illumina 16S rRNA gene sequencing during the operating period (30

95

time points) at two different sampling locations in the bioreactor and correlated the relative

96

abundance data with n-caprylate specificity data.

97 98

MATERIALS AND METHODS

99

Growth Medium, Inoculum, and Bioreactor Setup

100

The composition for the growth medium that we used has been described previously.14, 15 For

101

each operating period, the substrate concentrations of ethanol and acetate in the growth medium

102

were varied to achieve the targeted ethanol-to-acetate substrate ratios (Table 1). At the end of a

103

previous study in our lab,15 the bioreactor had been overloaded for research purposes, and as a

104

result its performance had suffered. We kept the reactor microbiome in the system, however, and

105

between the end of the previous study and the beginning of Period 1 an operating period of

5 ACS Paragon Plus Environment

Environmental Science & Technology

106

approximately six months was required to recover the performance (data not shown). The

107

bioreactor system was described previously.15 An upflow anaerobic filter (working volume 0.7

108

L) was operated with a continuous in-line membrane-based liquid-liquid extraction (i.e.,

109

pertraction) system. The substrate feed flow rate that we used was approximately 0.6 L d-1, while

110

the system recycle flow rate was 130 L d-1, which resulted in a recycle-to-feed ratio of ~220. The

111

hydraulic retention time (HRT) was ~1.2 days. The temperature of the bioreactor was maintained

112

at 30±1ºC, while the pH of the bioreactor broth was maintained at 5.26±0.09 via addition of 0.5

113

M hydrochloric acid to the well-mixed recycle and feed inlet at the base of the bioreactor. The

114

alkaline extraction solution was initially buffered with 0.3 M sodium borate and was then

115

maintained at pH 9.48±0.34 via addition of 5 M sodium hydroxide. When the n-caprylate

116

concentration in the alkaline extraction solution reached ~150-200 mM, two-thirds of the

117

extraction solution was removed (so 2-L out of ~ 3-L) and deionized water plus 5 M sodium

118

hydroxide was used to bring it back to its original 3-L volume. The purpose of this step was to

119

maintain efficient extraction rates from the bioreactor by maintaining low n-caprylate

120

concentrations in the extraction solution.

121 122

Operating Periods

123

Information on analysis and calculations is given in the supplementary information. When we

124

use the terminology ethanol-to-acetate substrate ratio, we consistently report the ratio of the

125

measured ethanol and acetate concentrations in the influent and express this ratio as g COD g

126

COD-1, with the molar ratio included in parenthesis for convenience. Measured moles of ethanol

127

and carboxylates were converted to a g COD basis using the following conversion factors (g

128

COD to mole): 96 (ethanol); 64 (acetate); 112 (propionate); 160 (n-butyrate), 208 (n-valerate),

6 ACS Paragon Plus Environment

Page 6 of 32

Page 7 of 32

Environmental Science & Technology

129

256 (n-caproate), 304 (n-heptanoate), and 352 (n-caprylate). On the other hand, when we use the

130

terminology ethanol-to-acetate liquor ratios, this is in relationship to our thermodynamic

131

model for which we used the ratio of the measured concentrations of ethanol and acetate (mM) in

132

the effluent, which we assume the active biomass to experience in the mixed liquor. During the

133

acclimation period (Period 1), the average organic loading rate (OLR) was 18.7 g COD L-1 d-1

134

with an ethanol-to-acetate substrate ratio of 11.7 g COD g COD-1 (7.8 mole mole-1) (Table 1).

135

During the main period (Periods 2 to 7), we operated the bioreactor at the following ethanol-to-

136

acetate substrate ratios: 275 (Period 2), 16.9 (Period 3), 6.7 (Period 4), 3.6 (Period 5), 2.9 (Period

137

6), and 1.8 g COD g COD-1 (Period 7) (equivalent molar ratios are: 183, 11.3, 4.5, 2.4, 1.9, and

138

1.2 mole mole-1, respectively), while maintaining a higher OLR of between 23.3 to 28.2 g COD

139

L-1 d-1 compared to Period 1 (Table 1). We operated the bioreactor at each ratio for an operating

140

period of at least two weeks (11 HRTs). During the recovery period (Periods 8 to 11), following

141

a loss of performance at an ethanol-to-acetate substrate ratio of 1.8 g COD g COD-1 (i.e., no

142

detectable n-caprylate production and a decrease in n-caproate production rates, Table S2), we

143

ran another set of ethanol-to-acetate substrate ratios: 5.1 (Period 8), 6.5 (Period 9), 6.0 (Period

144

10), and 135 g COD g COD-1 (Period 11) (equivalent molar ratios are: 3.4, 4.3, 4, and 90 mole

145

mole-1, respectively) (Table 1).

146 147 148 149 150 151

Table 1. Operating data for each period in the acclimation, main, and recovery periods are reported as mean ± standard error. During Period 9, the extraction system was switched off for the bioreactor. For Period 10, measurements of acetate were not available (na) for the influent. Therefore, values reported for ethanol-to-acetate substrate ratio and OLR are an approximation. Period

Days

HRT (d)

Ethanol (mM)

Acetate (mM)

OLR (g COD L-1 d-1)

1.2 ± 0.02

Ethanol-to-acetate substrate ratio (g COD g COD-1) 11.7 ± 1.0

Period 1

0 to 42

200 ± 9.42

25.7 ± 1.75

18.7 ± 0.81

Period 2

43 to 55

1.17 ± 0.05

275 ± 58.3

289 ± 3.64

1.58 ± 0.33

25.2 ± 1.1

Period 3

56 to 71

1.27 ± 0.07

16.9 ± 2.1

282 ± 5.21

25.0 ± 3.01

23.9 ± 1.43

7 ACS Paragon Plus Environment

Environmental Science & Technology

Page 8 of 32

Period 4

72 to 84

1.16 ± 0.04

6.7 ± 0.4

282 ± 10.5

63.4 ± 3.35

28.2 ± 1.24

Period 5

85 to 101

1.17 ± 0.04

3.6 ± 0.3

223 ± 10.65

94.1 ± 4.74

25.2 ± 1.18

Period 6

102 to 119

1.2 ± 0.05

2.9 ± 0.1

216 ± 5.26

112 ± 3.41

24.6 ± 1.08

Period 7

120 to 136

1.2 ± 0.04

1.8 ± 0.1

178 ± 4.53

146 ± 6.24

23.3 ± 0.85

Period 8

137 to 181

1.14 ± 0.04

5.1 ± 0.5

218 ± 9.68

64.6 ± 6.1

23.5 ± 1.16

Period 9

182 to 202

1.2 ± 0.04

6.5 ± 0.5

245 ± 11.92

56.4 ± 3.55

24.0 ± 1.31

Period 10

203 to 214

1.14 ± 0.04

6

221 ± 17.14

na

24

Period 11

215 to 242

1.12 ± 0.05

135 ± 18.1

287 ± 6.08

3.18 ± 0.42

26.2 ± 1.35

152 153

Microbial Community Analysis

154

We collected biomass samples for Illumina 16S rRNA gene sequencing analysis from ports

155

located at the bottom and middle of the bioreactor height approximately weekly throughout the

156

operating period by following the method previously described by Kucek et al.15 The bioreactor

157

was mixed by withdrawing and refilling a 60-mL syringe ten times at each sampling port to

158

resuspend the flocculent biomass prior to sampling. We did not observe clearly defined biofilms

159

on the packing material in the bioreactor, so we did not perform biofilm sampling. Pelleted

160

biomass samples were stored at -80°C until further processing. Genomic DNA was extracted

161

using the PowerSoil-htp 96 Well Soil DNA Isolation kit (MO BIO Laboratories Inc., Carlsbad,

162

CA) according to the protocol of the manufacturer. The DNA amplification protocol was

163

described previously31 with the following changes: Mag-Bind RxnPure Plus magnetic beads

164

solution (Omega Biotek, Norcross, GA, USA) was used instead of Mag-Bind E-Z Pure and 50 ng

165

DNA per sample was pooled instead of 100 ng. Duplicate PCR reactions of each DNA extract

166

were performed and pooled prior to sequencing. Paired-end reads were joined in QIIME

167

v.1.9.132 using the joined_paired_ends.py script and then the joined reads were uploaded to

168

QIITA (qiita.microbio.me) for further processing. The sortmerna method33 was used to bin

169

sequences in operational taxonomic units (OTUs) at 97% identity. Taxonomy was assigned for

170

representative sequences selected for each OTU using the Greengenes v.13.8 database.34 The 8 ACS Paragon Plus Environment

Page 9 of 32

Environmental Science & Technology

171

remaining analyses were performed locally in QIIME and R v.3.3.2.35 Singletons (i.e., OTUs that

172

were observed fewer than two times in the entire dataset) were removed from the dataset

173

resulting in 932 unique OTUs.

174 175

Alpha diversity was analyzed via the Gini coefficient (a measure of unevenness), observed OTUs

176

(richness), and Shannon diversity36 metrics available in QIIME. One hundred rarefactions at a

177

depth of 6,510 sequences per sample were performed and collated. Statistical analyses of the

178

alpha diversity results were performed using analysis of variance (ANOVA) and Tukey HSD in

179

R. In addition, heat maps were created to represent OTU relative abundance via the gplots

180

package in R.37 Correlations of OTU relative abundance with n-caprylate specificities was

181

investigated using the Spearman’s rank correlation coefficient via rcorr function38,

182

Hmisc package in R. P-values were adjusted for multiple comparisons using the false discovery

183

rate (FDR) method40 via the function p.adjust in the stats package in R. Correlations with

184

adjusted p-values |ΔGreq| for scenario 1 in Fig. S2), as well as the formation of mixtures of n-butyrate, n-

327

caproate, and n-caprylate (|ΔGrxn|>|ΔGreq| for scenarios 2 and 3 in Fig. S2). Indeed, we observed

328

high n-caprylate-to-n-caproate product ratios of 5.9 and 2.0 g COD g COD-1 (4.3 and 1.5 mole

329

mole-1) and n-caprylate specificities of 82% and 64% for our bioreactor during Periods 2 and 3,

330

respectively (Table S1).

331 332

During Period 4 at an ethanol-to-acetate substrate ratio of 6.7 g COD g COD-1 (4.5 mole mole-1)

333

(Table 1), the ethanol-to-acetate liquor ratio declined to 3.11 (Table S3). At this ratio that was

334

observed by the microbiome, the model predicted that formation of only n-caprylate was not

335

thermodynamically feasible, but formation of a mixture of n-butyrate, n-caproate, and n-

336

caprylate was (|ΔGrxn|>|ΔGreq| for scenarios 2 and 3, but not scenario 1 in Fig. S2). This mixed

337

product spectrum was experimentally verified (Fig. 1 and Table S1). At the lowest ethanol-to-

338

acetate substrate ratios that we maintained during the main period (i.e., 3.6, 2.9 and 1.8 during

339

Periods 5, 6, and 7, respectively, Table 1), the ethanol-to-acetate liquor ratios were also

16 ACS Paragon Plus Environment

Page 16 of 32

Page 17 of 32

Environmental Science & Technology

340

correspondingly low (i.e., 1.4, 1.0, and 0.6, respectively, Table S3). At these liquor ratios none

341

of the scenarios that we tested were thermodynamically feasible (Fig. S2), ruling out

342

considerable amounts of n-caprylate production from our thermodynamic model. However, we

343

did observe lower n-caprylate production during Periods 5 and 6 with specificities of 25.7% and

344

23.7%, respectively, but did not observe any n-caprylate production during Period 7 (Fig. 1 and

345

Table S1). These experimental specificities for Periods 5 to 6 were, indeed, lower than the

346

alleged specificity for our 3rd scenario (i.e., 1.5 moles n-butyrate, 1.5 moles n-caproate, and

347

0.625 moles n-caprylate), which would have translated into a 35.3% n-caprylate specificity based

348

on COD, and which was not possible according to our model (Fig. S2).

349 350

Even though the predictions of our model are in agreement with the bioreactor performance, real

351

care should be taken in using the model, because of our over-simplifications and assumptions.

352

First, we assumed that the effluent ethanol and acetate concentrations represented a uniform

353

concentration in the bioreactor experienced by the microbiome. In fact, we saw different

354

microbiome compositions at two different sampling locations in our bioreactor across the study

355

period (discussed later in the text), which suggests that the substrate composition (i.e., ethanol-

356

to-acetate liquor ratio may have varied throughout the bioreactor height. Second, we developed a

357

formula for the moles of ATP produced by our model assuming that the Rnf complex

358

(ferredoxin-NAD oxidoreductase complex) found in C. kluyveri was being used to establish a

359

proton gradient across the cell membrane and drive ATP production. Wang et al.42 have reported

360

that under low ethanol and acetate concentrations (~1 mM), and therefore correspondingly lower

361

reaction Gibbs free energies, C. kluyveri can adjust the activity of two enzymes (an electron

362

bifurcating Nfn complex and a NADP+dependent β-hydroxybutyryl-CoA dehydrogenase) to

17 ACS Paragon Plus Environment

Environmental Science & Technology

363

decrease its ATP production requirements. Our current model does not account for this potential

364

shift in ATP production (and lower associated Gibbs free energy requirements) in considering

365

the thermodynamic feasibility of n-caprylate production. Third, our model was based on the

366

reported metabolism of C. kluyveri, which may not represent the actual situation for our reactor

367

microbiome. Fourth, this model was designed for a system with product extraction (e.g., our

368

bioreactor) where n-caprylate is continuously being removed from the bioreactor broth. Our

369

model does not account for the effect of product toxicity on n-caprylate production, and therefore

370

cannot be directly applied to batch systems without product extraction.

371 372

Primarily Ethanol in the Substrate Led to Lower MCC Production Rates Than When

373

Acetate was Present

374

To our knowledge, this is the first study to demonstrate n-caproate and n-caprylate production

375

from primarily ethanol as the substrate of a bioreactor (i.e., an ethanol-to-acetate ratio of 275 g

376

COD g COD-1 [183 mole mole-1] during Period 2). Even though ethanol as the primary substrate

377

led to the highest n-caprylate specificity of 81.9% (Table S1), it also led to a lower sum of n-

378

butyrate and MCC production rates (to estimate the total chain elongation rates) of 13.2 g COD

379

L-1 d-1 during Period 2 compared to 17.6 g COD L-1 d-1 during Period 1 (Table S2) at a

380

considerably higher OLR of 25.2 compared to 18.7 g COD L-1 d-1, respectively (Table 1).

381

During Periods 3 to 6 with similar OLRs than Period 2, the sum of n-butyrate and MCC

382

production rates was also higher compared to Period 2 and remained between 14.5 and 19.9 g

383

COD L-1 d-1 (Table S2), resulting in higher n-butyrate plus MCC conversion efficiencies during

384

Periods 3 to 6 compared to during Period 2 (Table S4). These conversion efficiencies increased

385

from 60.5% and 80.6% between Period 3 and Period 6, while the conversion efficiency was

18 ACS Paragon Plus Environment

Page 18 of 32

Page 19 of 32

Environmental Science & Technology

386

52.6% during Period 2 (Table S4). Thus, the total chain elongation rate was considerably lower

387

for Period 2 when primarily ethanol was used as a substrate compared to the period before and

388

after. The lower sum of n-butyrate and MCC production rates resulted in a higher average

389

effluent ethanol concentration during Period 2 compared to Periods 3 to 6 with a maximum

390

concentration up to 95 mM on Day 61 during Period 2 (Fig. 2B), even though the OLR was

391

similar (Table 1). Therefore, ethanol availability was not the reason for the sluggish chain

392

elongation rates. The ethanol concentration remained well below the concentrations that have

393

previously been found to be inhibitory in a pure-culture study (460 mM)26 and in an open-culture

394

study (~300 mM).15

395 396

Our model does not provide kinetic information, but it may guide us in finding an explanation

397

for the sluggish chain elongation rates during Period 2. At the highest ethanol-to-acetate ratios,

398

our thermodynamic model (Eq. 1) predicted an increased production of hydrogen via reverse β-

399

oxidation. Possibly, increased hydrogen concentrations in the bioreactor at the highest ethanol-

400

to-acetate ratio acted to slow the kinetics by lowering the ATP production of reverse β-oxidation,

401

and therefore by reducing the n-caprylate production rates. This was explained before by

402

Angenent et al.20 because of a shift from a pathway with both substrate-level phosphorylation

403

and transport-coupled phosphorylation at low hydrogen partial pressures to a pathway with

404

solely substrate-level phosphorylation pathway at high hydrogen partial pressures. Here, we have

405

not been able to experimentally confirm this hypothesized shift in pathways. Others found hints

406

as well for a slower kinetic rate of chain elongation with ethanol when hydrogen was anticipated

407

to be at higher concentrations. For example, Roghair et al.43 found slow kinetic rates for chain

408

elongation with ethanol in an ethanol/propionate fed bioreactor, but also found that by adding

19 ACS Paragon Plus Environment

Environmental Science & Technology

409

CO2 to lower the hydrogen partial pressure (due to methanogenesis), the rate of chain elongation

410

with ethanol was stimulated greatly. A different explanation for a more sluggish chain elongation

411

rate with increasing ethanol-to-acetate molar ratios was given by Leng et al.,44 who found an

412

optimum energy release at a ratio of 3 with a simple thermodynamic model that included only n-

413

butyrate and n-caproate production.

414 415

Regardless of the explanation for a slower chain-elongating rate with primarily ethanol, we did

416

not find a clear optimum ethanol-to-acetate substrate ratio that would satisfy the maximum n-

417

caprylate specificity, the n-caprylate production rate, and the total chain elongation rate. Clearly,

418

the highest n-caprylate specificity of 82% was observed with the highest ethanol-to-acetate

419

substrate ratio of 275. However, primarily feeding ethanol at this ratio decreased the n-caprylate

420

production rates and also the total chain elongation rates, and thus care should be taken to not

421

over load the system with such lower production rates by lowering the organic loading rates. We

422

found the total chain elongation rate to be the highest at 19.9 g COD L-1 d-1 at an ethanol-to-

423

acetate substrate ratio of 2.9 g COD g COD-1 (1.9 mole mole-1) during Period 6 (Table S2).

424 425

Reactor Microbiomes in Bottom and Middle of Bioreactor Were Different

426

We performed a time-series analysis of the microbiome sampled from the bottom and middle

427

sampling ports of our upflow bioreactor during all the study periods (Periods 1 to 11). Clear

428

differences in the composition of the microbiomes from the bottom and middle port of the

429

bioreactor were observed throughout the operating period (Fig. 3, Fig. S3, Fig. S4). The majority

430

of samples from the bottom of the bioreactor had a higher relative abundance of the phylum

431

Firmicutes compared to the phylum Proteobacteria (Fig. S4A), whereas the reverse was true in

20 ACS Paragon Plus Environment

Page 20 of 32

Page 21 of 32

Environmental Science & Technology

432

the middle of the bioreactor (Fig. S4B). The bottom and middle port samples consisted of 53 and

433

56 OTUs, respectively, which reached a higher than one percent relative abundance in at least

434

one of the bioreactor samples (Fig. 3 and Fig. S3). Of these OTUs, the bottom port had 15

435

unique OTUs (primarily belong to the phylum Firmicutes, highlighted in blue in Fig. 3) and the

436

middle port had 18 unique OTUs (primarily belonging to the phylum Proteobacteria, highlighted

437

in blue in Fig. S3). This difference in communities is similar to the microbial community

438

stratification that has previously been observed in upflow anaerobic sludge blanket reactors for

439

methane production.45 Potential differences in substrate concentrations, pH, and oxygen

440

concentrations throughout the height of the bioreactor may have played an important role in the

441

differences observed. Though we employed a high recycle-to-feed flow rate ratio (~220) in this

442

study, the packing material in the bioreactor may have interfered with efficient mixing of the

443

bioreactor. From anaerobic digestion studies, we already know that the composition of substrate

444

is a large driver of the community composition46. Because we were adding our substrate to the

445

base of the bioreactor, we hypothesize that a higher chain-elongating activity occurred in the

446

bottom of the bioreactor compared to in the middle. Though, more work would need to be

447

performed to confirm this. Our observations of different microbial communities in the two

448

sampling locations highlight the need for multiple microbiome sampling locations in a bioreactor

449

that is not completely mixed. We note that the study by Kucek et al.15 only characterized the

450

microbiome in the middle of the bioreactor.

451 452

Despite the differences observed in our microbiomes, we did see some common trends. We

453

examined which populations of OTUs in the bottom and middle ports of our bioreactor were

454

significantly positively or negatively correlated with n-caprylate specificities (FDR-adjusted p-

21 ACS Paragon Plus Environment

Environmental Science & Technology

455

value