Controlling Anaerobic Digestion to Produce Targeted Compounds

Controlling anaerobic digestion to produce targeted compounds

Miriam Peces Gomez

Bachelor in Chemical Engineering Master in Environmental Engineering

A thesis submitted for the degree of Doctor of Philosophy at The University of Queensland in 2017 School of Civil Engineering Centre for Solid Waste Bioprocessing

Abstract Anaerobic digestion is a mix-culture microbial-mediated process that has primarily been applied to produce methane and stabilise organic matter. However, the intermediates of the digestion process (volatile fatty acids, alcohols, and hydrogen) have applications as commodity chemicals or as precursors to a range of biobased products. However, one of the main challenges to broaden the application of anaerobic digestion is the difficulty associated with robustly controlling mixed-culture products, such that a suite of products can be repeatedly produced. Therefore, understanding how the change in microbial populations or loss in microbial functionality influence the behaviour of the rest of the community can prove to be a powerful tool for manipulating and controlling processes towards a desired commodity.

The impact of the starting inoculum on long-term anaerobic digestion performance, metabolic activity rates and microbial community composition remains unclear. To understand the impact of starting inoculum, active microbial communities from four different full-scale anaerobic digesters were each used to inoculate four continuous anaerobic digesters. Thereafter, the digesters were -1 -1 operated identically at 15 days solid retention time, an organic loading rate of 1 g COD Lr d (75:25 - cellulose:casein), and 37 ºC for 295 days. The digesters performance converged and stabilised in 80 days, while activity rates and microbial communities converged and stabilised after 145 days of operation. After 295 days, 52% of all identified OTUs were common to all digesters, and this core community accounted for 72% of the total microbial community relative abundance defined by various bacterial taxa (Bacteroidales, Ruminoccocaceae, Kosmotoga and Treponema) and archaeal taxa (Methanosaeta, Candidatus Methanoregula and Methanospirillum). This indicates that deterministic factors (process operational conditions) were a stronger driver in controlling the ultimate microbial composition in a digester rather than the initial microbial community composition. Moreover, Pearson correlation coefficients revealed several significant associations between bacterial taxa found in the digesters and activity rate profiles. For instance, the presence of Armatimonadetes was positively correlated with higher cellulolytic rates and bacteria belonging to genus Synthophobacter and Clostridum or families Veillonellaceae and candidate BA008 (phylum Bacteroidetes) were correlated to higher butyrate and propionate degradation rates. Overall, it seems plausible that process operational conditions can be used to tune microbial composition and functionality in an anaerobic digester.

To explore the extent that the anaerobic digestion process can be manipulated by a sole selection pressure, the solid retention time was isolated as a pressure parameter. Without interruption, the same four continuous anaerobic digesters were subjected to a sequential decrease in solid retention time -1 -1 from 15 to 8 to 4 to 2 days while maintaining the organic loading rate at 1 g COD Lr d , the same

i substrate composition ratio (75:25 - cellulose:casein) and the same temperature (37 ºC). Each solid retention time was operated until steady state was achieved. Results showed that acetoclastic methanogenesis carried out by Methanosaeta remained active down to 2 day solid retention time and only minor accumulation of volatile fatty acids was achieved (less than 3.5% of influent COD). Therefore, solid retention time as an individual selection pressure was not an effective parameter to shift the anaerobic digestion product profile. However, lowering solid retention times induced a shift in metabolic activity rates, where ethanol degradation gained dominance over butyrate and propionate degradation. Solid retention time also influenced the microbial dynamics of the digesters, driving changes at family or genus level, although the most noticeable finding was the formation of biofilms containing a high abundance of Methanosaeta at the lowest solid retention time. This suggests that the different microbial communities in all four digesters developed similar survival strategies under non-favourable methanogenic conditions.

To contextualise and prove the applicability of imposed conditions to steer the process, a combination of temperature, retention time and oxygen availability were selected to control the fermentation patterns of primary sludge followed by anaerobic digestion to recover biogas, as part of a bio-refinery concept. Primary sludge pre-fermentation was carried out at different temperatures (20, 37, 55, 70ºC), treatment times (12, 24, 48, 72h), and oxygen availability (semi-aerobic, anaerobic). pH was not controlled. The anaerobic biodegradability after pre-fermentation was evaluated using biochemical methane potential tests. The results showed that fermentation at 20 and 37 ºC was optimal for volatile fatty acids production with acetate and propionate being major products. Anaerobic fermentation at 37, 55 and 70 ºC resulted in higher solubilisation yield at the expense of reduced methane production by 20%, while semi-aerobic fermentation allowed both volatile fatty acids recovery and improved methane potential. Replication experiments using a different batch of primary sludge showed that the main trends could be reproduced exemplifying that fermentation and anaerobic digestion products can be controlled by operational decisions.

ii Declaration by author

This thesis is composed of my original work, and contains no material previously published or written by another person except where due reference has been made in the text. I have clearly stated the contribution by others to jointly-authored works that I have included in my thesis.

I have clearly stated the contribution of others to my thesis as a whole, including statistical assistance, survey design, data analysis, significant technical procedures, professional editorial advice, and any other original research work used or reported in my thesis. The content of my thesis is the result of work I have carried out since the commencement of my research higher degree candidature and does not include a substantial part of work that has been submitted to qualify for the award of any other degree or diploma in any university or other tertiary institution. I have clearly stated which parts of my thesis, if any, have been submitted to qualify for another award.

I acknowledge that an electronic copy of my thesis must be lodged with the University Library and, subject to the policy and procedures of The University of Queensland, the thesis be made available for research and study in accordance with the Copyright Act 1968 unless a period of embargo has been approved by the Dean of the Graduate School.

I acknowledge that copyright of all material contained in my thesis resides with the copyright holder(s) of that material. Where appropriate I have obtained copyright permission from the copyright holder to reproduce material in this thesis.

iii Publications during candidature

Peer-reviewed journal papers

Puyol, D., Batstone, D.J., Hülsen, T., Astals, S., Peces, M., Krömer, J.O. 2017. Resource recovery from wastewater by biological technologies: Opportunities, challenges, and prospects. Frontiers in Microbiology, 7, article 2106.

Peces, M., Astals, S., Clarke, W.P. and Jensen, P.D., 2016. Semi-aerobic fermentation as a novel pre-treatment to obtain VFA and increase methane yield from primary sludge. Bioresource technology, 200, pp.631-638.

Nolla-Ardèvol, V., Peces, M., Strous, M., Tegetmeyer, H.E., 2015. Metagenome from a Spirulina digesting biogas reactor: Analysis via binning of contigs and classification of short reads. BMC Microbiology, 15-1, article 15.

Peces, M., Astals, S., Mata-Alvarez, J., 2015. Effect of moisture on pretreatment efficiency for anaerobic digestion of lignocellulosic substrates. Waste Management, 46, pp.189-196.

Conference papers

Peces, M., Jensen, P.D., Astals, S., and Clarke, W.P. 2016. Do different inocula converge given the same operational conditions in long-term anaerobic digestion? In proceedings: XII DAAL - Taller y Simposio Latino Americano en Digestión Anaerobia. Cusco, Peru. (Conference oral presentation)

Peces, M., Jensen P.D., Clarke, W.P. and Astals, S., 2015. Semi-aerobic pre-fermentation conditions to recover VFA and improve methane potential. In proceedings: 14th World Congress on Anaerobic Digestion. Viña del Mar, Chile. (Conference oral presentation)

iv Publications included in this thesis

This paper has been modified an incorporated as Chapter 7. Contributor Statement of contribution Miriam Peces (Candidate) Designed experiments (30%) Conducted experiments (80%) Data analysis (90%) Wrote the paper (60%) Sergi Astals Designed experiments (70%) Conducted experiments (20%) Data analysis (10%) Wrote the paper (30%) Critically reviewed and edited the paper (40%) William P. Clarke Wrote paper (5%) Critically reviewed and edited the paper (20%) Paul D. Jensen Wrote paper (5%) Critically reviewed and edited the paper (40%)

v Contributions by others to the thesis “No contributions by others.”

Statement of parts of the thesis submitted to qualify for the award of another degree “None”.

vi Acknowledgements This PhD has been a journey in many ways, and it would not have been possible without the support, guidance and encouragement from many people.

I would like to start extending my gratitude to my advisors Prof William Clarke and Dr Paul Jensen. Bill, thank you very much for giving me the opportunity to pursue my PhD allowing me the freedom to explore my own paths being so supportive along the way. I truly admire your patience, optimism and the care you’ve always shown for your students. Paul, I’d always be grateful for lining up this opportunity and happy that you accepted to be involved in the advisory team sharing your knowledge and bringing up stimulating debates.

Special thanks to Prof Damien Batstone and Prof Zhiguo Yuan for being part of the reviewing committee, keeping track of this candidature and foremost giving the time for providing insightful comments and recommendations. I would also like to acknowledge the financial support received from the UQI scholarship and the funding provided by the Centre for Solid Waste Bioprocessing to carry out this research.

But this journey started a long time ago at the University of Barcelona. I cannot be grateful enough to Prof Joan Mata-Alvarez who gave me the opportunity to work in the Environmental Biotechnology group, directing my BSc and MSc theses, and kindly encouraged me to pursue the PhD in the other side of the world.

To my mentors Dr Sergi Astals and Dr Yang (Kenn) Lu. Sergi, thank you for your excellent mentorship since the very beginning of this journey sharing your talent, experience, enthusiasm and ideas. Your passion for research has definitely inspired many to follow the research path. Kenn, you were like a magician to me. Million thanks for introducing me to the techniques to explore the fascinating world of microbes, sharing the tips and tricks that turned magic into science and helping me anytime I run aground.

Like any other journey, many people are involved keeping the practical things running smoothly and worry-free. Thanks to Dr Beatrice Keller-Lehmann, Nathan Clayton and Dr Katrin Sturm from the Advanced Water Management Centre analytical services lab; and Dr Nicola Angel from the Australian Centre for Ecogenomics for performing the 16S rRNA Illumina pyrosequencing. Many thanks to the School of Civil admin team for their day-to-day work and especially to Wendy for making the finance paperwork a friendly task.

vii Sometimes journeys take parallel paths and although not part of this dissertation, I would like to thank Dr Denys Villa for inviting me to take part in several of her projects acquiring new skills, and the students we co-mentored, Amy and Ben. It has been a very fun and fulfilling experience.

I am indebted to Miheka Patel and Lizanne Obersky, you have been the best travelling companions I could have ever imagined to have. I guess it is mandatory to start by the formal gratitude. Thanks for looking after my four little monsters in my absence and making the office/lab a more than pleasant place to work in, but it really goes far beyond that. I will not forget all the coffees, wine, cheese, laughs, gossips and confidences that have made this journey a truly enjoyable experience and I can’t summarise here. And the fantastic four would not be complete without Julijana, the youngest spirit of the team and chocolate supplier, I wonder what would be of us without you! I am looking forward to the new adventures to come.

I would like to thank the awesome people with whom I shared many memorable moments in and out uni. Apra, Babet, Bea, Elnaz, Emma, Federica, Guillermo, Heidy, Julia, Justus, Katie, Ludwika, Mike, Miriam, Shao and Rob thank you very much for not letting me miss home!

Last but not least, thanks to my family and husband for their unconditional love and support. Mama, papa, y iaia gracias por los buenos valores que me habéis enseñado que ningún título te puede ofrecer. Laia, no tengo palabras, gracias por apoyarme y agunatarme durante horas hablando de la tesis. Porque pese a irme a las antípodas nunca habéis estado lejos, como dice la iaia “fíjate que cosas, ahí tan lejos y lo bien se la oye”. Finalment a tu, Sergi. Gràcies per confiar en mi, animar-me i estimar- me. Sóc extremadament afortunada de tenir algú com tu com a company de viatge. Al teu costat no em fa por afrontar els reptes i espero que continuem viatjant junts, de la mà, allà cap a on ens dugui el vent.

Thank you all!

viii Keywords Anaerobic digestion, fermentation, microbial dynamics, inocula, deterministic, functionality, kinetic rates, correlation, retention time, multivariate analyses.

Australian and New Zealand Standard Research Classifications (ANZSRC) ANZSRC code: 090703, Environmental Technologies, 40% ANZSRC code: 060504, Microbial Ecology, 35% ANZSRC code: 090409 Wastewater Treatment Processes, 25%

Fields of Research (FoR) Classification FoR code: 0907, Environmental Engineering, 40% FoR code: 0605, Microbiology, 35% FoR code: 0904, Chemical Engineering, 25%

ix Table of Contents Abstract ...... i Declaration by author ...... iii Publications during candidature ...... iv Peer-reviewed journal papers ...... iv Conference papers ...... iv Publications included in this thesis ...... v Contributions by others to the thesis ...... vi Statement of parts of the thesis submitted to qualify for the award of another degree ...... vi Acknowledgements ...... vii Keywords ...... ix Australian and New Zealand Standard Research Classifications (ANZSRC) ...... ix Fields of Research (FoR) Classification ...... ix Table of Contents ...... x List of Figures ...... xiv List of Tables ...... xix List of abbreviations ...... xx 1 Introduction ...... 1 1.1 Project Significance ...... 1 1.2 Project rationale and methodological approach ...... 2 2 Literature Review ...... 4 2.1 Anaerobic Digestion ...... 4 2.1.1 Hydrolysis ...... 5 2.1.2 Acidogenesis / Primary Fermentation ...... 7 2.1.3 Acetogenesis / Secondary Fermentation ...... 10 2.1.4 Methanogenesis ...... 11 2.2 Operational parameters ...... 12 2.2.1 Inoculum ...... 12 2.2.2 Substrate characteristics ...... 14 2.2.3 Temperature ...... 15 2.2.4 Retention time and organic loading rate ...... 18 2.2.5 pH ...... 20 2.3 Linking microbes and process: Molecular tools and statistical methods ...... 20 2.3.1 Phylogenetic characterisation ...... 21 2.3.2 Functional capacity and expression: Advanced molecular techniques ...... 22 2.3.3 Microbial to process. Integrated approaches and statistical methods …………………...25 2.3.3 Microbial to process. Integrated approaches and statistical methods ...... 25 3 Research gaps and objectives ...... 26

x 3.1 Research gaps ...... 26 3.2 Research objectives ...... 27 4 Materials and methods ...... 28 4.1 Inoculum sources and substrates ...... 28 4.1.1 Inocula ...... 28 4.1.2 Substrates ...... 29 4.1.2.1 Digesters substrates and media ...... 29 4.1.2.2 Activity assays substrates ...... 29 4.1.2.3 Fermentation assays substrates ...... 30 4.2 Experimental Set up ...... 30 4.2.1 Continuous digesters ...... 30 4.2.1.1 Determine influence of inoculum source ...... 31 4.2.1.2 Determine the influence of retention time ...... 31 4.1.2 Activity assays ...... 32 4.2.3 Fermentation batch assays ...... 33 4.2.4 Biochemical methane potential test ...... 33 4.3 Methods ...... 34 4.3.1 Analytical methods ...... 34 4.3.1.2 Physico-chemical analyses ...... 34 4.3.1.2 Microbial analyses ...... 34 4.3.2 Data Analysis ...... 35 4.3.2.1 Performance-related calculations ...... 35 4.3.2.2 Activity assays and BMP modelling ...... 36 4.3.2.4 Statistical analyses ...... 38 5 Deterministic mechanisms define anaerobic digestion microbiome and its functionality regardless of the initial microbial community ...... 39 5.1 Introduction ...... 40 5.2 Aim and approach ...... 41 5.3 Results ...... 42 5.3.1 Digester operation ...... 42 5.3.2 Activity assays ...... 43 5.3.3 Microbial communities ...... 45 5.3.4 Linkage of microbial composition to functionality ...... 47 5.4 Discussion ...... 49 5.4.1 Microbial composition linked to process function and activity rates ...... 49 5.4.1.1 Process and microbial community performance...... 49 5.4.1.2 Individual microbial populations associated with activity rates ...... 50 5.4.2 Development of microbial communities towards a core-community ...... 51 5.5 Conclusions ...... 52

xi 6 Transition of microbial communities and degradation pathways in anaerobic digestion at decreasing retention time ...... 53 6.1 Introduction ...... 54 6.2 Aim and approach ...... 55 6.3 Results ...... 56 6.3.1 Digester operation ...... 56 6.3.2 Activity assays ...... 58 6.3.3 Microbial communities ...... 59 6.3.3.1 Influence of SRT ...... 59 6.3.3.2 Influence of the operational disturbance ...... 62 6.3.3.3 Biofilm formation ...... 62 6.4 Discussion ...... 63 6.4.1 Influence of SRT on AD process performance ...... 63 6.4.2 Activity rates inferred a switch in degradation pathways ...... 64 6.4.3 Influence of SRT on microbial communities ...... 65 6.4.4 Effect of an operational disturbance on microbial assembly ...... 66 6.5 Conclusions ...... 67 7 Semi-aerobic fermentation as a novel pre-treatment to obtain VFA and increase methane yield from primary sludge ...... 68 7.1 Introduction ...... 69 7.2 Aim and approach ...... 70 7.3 Results and Discussion ...... 71 7.3.1 Extraction of valuable compounds from primary sludge ...... 71 7.3.1.1 Organic matter solubilisation ...... 71 7.3.1.2 VFA distribution ...... 72 7.3.2 Extraction of soluble compounds and influence on PS methane yield ...... 75 7.4 Conclusions ...... 78 8 Conclusions and recommendations ...... 79 8.1 Overall conclusions ...... 79 8.2 Future directions ...... 81 References ...... 83 Appendix A: Supplementary material for Chapter 5 ...... 96 A.1 Digesters’ performance ...... 97 A.1.1 Individual VFA ...... 97 A.1.2 pH, tCOD, TS, and VS ...... 98 A.1.3 COD balance ...... 99 A.2 Relative metabolic rates ...... 100 A.3 Microbial communities ...... 101 A.3.1 NMDS and compositional dissimilarity based on Bray-Curtis distance ...... 101

xii A.3.2 Relative abundance explained by unique/shared OTUs over time ...... 102 A.3.3 Analysis of Procrustes rotations based on individual PCA ...... 103 A.3.4 Heatmap ...... 105 A.3.5 Correlation map metabolic rates-taxa ...... 106 Appendix B: Supplementary material for Chapter 6 ...... 108 B.1 Digesters’ performance ...... 109 B.1.1 Individual VFA yields ...... 109 B.1.1.1 Major VFA ...... 109 B.1.1.2 Minor VFA ...... 110 B.1.2 TS, VS, tCOD, sCOD, and pH ...... 111 B.1.3 Summary of operational conditions and process performance ...... 112 B.1.4 COD balance ...... 113 B.2 Relative metabolic rates ...... 114 B.3 Microbial communities ...... 116 B.3.1 PERMANOVA on PCA with SRT as a factor ...... 116 B.3.2 Individual PCA ...... 117 B.3.3 Bacterial and Archaeal Kingdom PCA ...... 118 B.3.4 Heatmap ...... 119 B.4 Biofilm pictures ...... 120 B.5 First-order fit cellulose hydrolysis and CSTR ...... 121 Appendix C: Supplementary material for Chapter 7 ...... 122 C.1 Cumulative methane production curves and confidence regions ...... 123 C.2 COD balance fermentation unit followed by anaerobic digestion ...... 126

xiii List of Figures Figure 1. Simplified scheme of the anaerobic digestion process. Adapted from Batstone et al. (2002) ...... 5 Figure 2. Simplified representation of cellulose hydrolysis mechanisms: (A) cellulosome system, (B) cell-associated system...... 6 Figure 3. Simplified scheme of major metabolic pathways for carbohydrate fermentation. Orange envelope represents a bacterial cell. Adapted from Hoelzle et al. (2014), Madigan et al. (2010), Temudo et al. (2007)...... 8 Figure 4. Scheme of major metabolic pathways for glucose fermentation by C. acetobutylicum, depending on the environmental pH. Adapted from Dürre (2005)...... 9 Figure 5. Example of a Stickland reaction for the amino acid pair Alanine - Glycine. Orange envelope represents a bacterial cell. Adapted from Madigan et al. (2010) ...... 10

Figure 6. VFA production and VFA distribution depending on substrate composition - 10 g CODinitial at 10 days. Adapted from Shen et al. (2014)...... 15 Figure 7. (A) Temperatures (minimum, maximum, and optimal) for microbial growth rate. (Adapted from (Madigan et al. 2010)). (B) Reaction rates of methane production from waste activated sludge at mesophilic range (Adapted from (Donoso-Bravo et al. 2009)) ...... 16 Figure 8. (A) VFA production and distribution and (B) DGGE gel band results for a mesophilic acidogenic digester running at 5, 3, and 2 SRT. Adapted from Maspolim et al. (2014) ...... 19 Figure 9. Combination of techniques to gain an understanding of bioprocesses ...... 25 Figure 10. Schematic representation of one digester set-up to evaluate the influence of inoculum source (4.2.1.1) and influence of retention time (4.2.1.2)...... 31 Figure 11. Schematic representation of the second experimental set-up to determine the influence of operational conditions into a bio-refinery concept...... 34 Figure 12. Schematic representation of the rationale and experimental set-up used in Chapter 5 ... 41 Figure 13. (A) Methane production yield of the 4 digesters (♦)SL, (♦)SS, (♦)PL, (♦)BG. Data within dotted red lines have been excluded from the data analysis due to operational complications. (B) VFA concentration. Operational stages correspond to the periods (■) Start-up, (■) Transition, and (■) Steady-State ...... 42

Figure 14. Evolution of activity rates (km) over time. The solid coloured line represents the substrate consumption rate for each digester (–) SL, (–) SS, (–) PL, (–) BG. Shadowed coloured box represents the 95% confidence interval of the activity rate value. (A) Cellulose hydrolysis activity, (B) butyrate activity, (C) propionate activity, (D) acetoclastic methanogenic activity, and (E) hydrogenotrophic methanogenic activity...... 44 Figure 15. PCA of the microbial community profiles at the OTU level (Hellinger transformed) for the four digesters over 10 sampling events (day 0 to 295). Each sample is represented by a single circle, coloured according to digester designations in previous plots. Circle size increases with time. OTUs are presented as black crosses, and the populations contributing most to the variability between microbial communities are identified. Correlation with performance parameters is indicated by the arrow...... 46 Figure 16. Microbial composition at 97% similarity. The heatmap shows the populations (>2.5% relative abundance in at least one sample) in the four digesters at the 10 sampling points. The lowest

xiv possible taxonomic assignment is shown in the right column and phylum level in the left column. Darker intensity indicates higher relative abundance; grey cells indicate that these taxa were not detected in the samples at the level of resolution (>0.01% of total counts at least in one sample). 47 Figure 17. Correlation maps display the taxa that showed a significant positive correlation with at least one metabolic rate. Left, Pearson correlation taxa- metabolism (linear). Right, Spearman correlation taxa-metabolism (monotonic). Blue represents positive correlations, red represents negative correlations with taxa abundance. Colour intensity represents the strength of the correlation (darker, stronger). Numbered cells indicate those where the correlation coefficient is significant (Bold, P <0.01, Italic P < 0.05)...... 48 Figure 18. Schematic representation of the rationale, experimental set-up used in Chapter 6 ...... 55 Figure 19. (A) Methane production yield of the 4 digesters (♦) SL, (♦) SS, (♦) PL, (♦) BG expressed in COD-equivalents (1g COD = 0.35 LNCH4). (B) Normalised VFA concentration (VFA yield). Data within dotted red lines have been excluded from the data analysis due to operational complications...... 57 Figure 20. Evolution of activity rates (km) over time, the solid coloured line represents the substrate consumption rate for each digester (–) SL, (–) SS, (–) PL, (–) BG. Shadowed coloured box represents the 95% confidence interval of the activity rate value...... 59 Figure 21. PCA of the microbial community profiles at the OTU level (Hellinger transformed) for the four digesters over 16 sampling events (4 at each SRT). Each sample is represented by a single circle, coloured according to digester designations in previous plots. Circle size decreases for each SRT. Squares represent the communities of the biofilms in each digester at the end of the experiment. OTUs are presented as black crosses, and the populations contributing most to the variability between microbial communities are identified. Correlation with performance parameters are indicated by arrows; blue arrows represent those parameters correlated with a significance lower than 0.01, and orange arrows parameters with a significance lower than 0.05 ...... 60 Figure 22. Microbial composition at 97% similarity. Heatmap shows the suspended biomass populations (>2 % relative abundance at least in one sample) in the four digesters at the 16 sampling points and the biofilm at the end of the experiment. The lowest possible taxonomic assignment is shown in the right column and phylum level in the left column. Darker intensity indicates higher relative abundance; grey cells indicate taxa not detected in the samples at the level of resolution (>0.005% of total counts at least in one sample) ...... 61 Figure 23. Schematic representation of the rationale and experimental set-up used in Chapter 7 ... 70 Figure 24. COD fractionation (in percentage) for each fermentation condition. (A) 20 ºC Semi- aerobic, (B) 20 ºC Semi-aerobic replicated, (C) 20 ºC Anaerobic, (D) 37 ºC Anaerobic, (E) 55 ºC Anaerobic, and (F) 70 ºC Anaerobic...... 71 Figure 25. VFA distribution depending of (A) 20 ºC Semi-Aerobic, (B) 20 ºC Semi-Aerobic (batch 2), (C) 20 ºC Anaerobic, (D) 37 ºC Anaerobic, (E) 55 ºC Anaerobic, and (F) 70 ºC Anaerobic, at the different exposure times applied...... 73 Figure 26. Cumulative specific methane production curves after fermentation at different temperatures, exposure time, and control. (A) 20 ºC Semi-Aerobic, (B) 20 ºC Semi-Aerobic (batch 2), (C) 20 ºC Anaerobic, (D) 37 ºC Anaerobic, (E) 55 ºC Anaerobic, and (F) 70 ºC Anaerobic. .. 75

xv Figure 27. Primary sludge after 72 h of fermentation (20 ºC Semi-Aerobic). (A) Graphic representation. (B) Photograph. 1: liquid rich in VFA; 2: residual sludge; and 3: semi-aerobic layer colonised by fungi...... 77 Figure A-I. Individual VFA concentration profiles of the 4 digesters (♦) SL, (♦) SS, (♦) PL, (♦) BG. (A) Acetic Acid, (B) Propionate acid, (C) Butyric acid, (D) Valeric acid. Operational stages correspond to the periods (■) Start-up, (■) Transition, and (■) Steady. Data within dotted red lines has been excluded from the data analysis due to operational complications ...... 97 Figure A-II. Monitoring parameters profiles of the 4 digesters (♦) SL, (♦) SS, (♦) PL, (♦) BG. (A) pH, (B) total COD, (C) total solids (TS), (D) volatile solids (VS). Operational stages correspond to the periods (■) Start-up, (■) Transition, and (■) Steady. Data within dotted red lines has been excluded from the data analysis due to operational complications ...... 98 Figure A-III. Evolution of the COD mas balance profiles of the 4 digesters (♦) SL, (♦) SS, (♦) PL, (♦) BG. Values close to 0 indicate that the theoretical COD removal calculated from the methane production equals to the experimental COD removal. Operational stages correspond to the periods (■) Start-up, (■) Transition, and (■) Steady. Data within dotted red lines has been excluded from the data analysis due to operational complications...... 99

Figure A-IV. Relative activity rates normalised by cellulose rate. The ratio km,i/km,cel >1 indicates cellulose hydrolysis is the limiting step. (A) Butyrate to cellulose, (B) Propionate to cellulose, (C) Acetate to cellulose, (D) Formate to cellulose. Solid coloured line represents ratio for each digester (–) SL, (–) SS, (–) PL, (–) BG. Shadowed coloured box represents the 95% confidence interval of the value...... 100 Figure A-V. (A) NMDS of the microbial community profiles at the OTU level (Bray-Curtis dissimilarity, stress = 0.171) for the four digesters over 10 sampling events (day 0 to 295). Each sample is represented by a single circle, coloured according to digester designations in previous plots. Circle size increases with time. (B) Compositional dissimilarity (Bray-Curtis distance) between the microbial communities in the four digesters at each time point relative to the initial community. (C) Compositional dissimilarity (Bray-Curtis distance) between the microbial communities in the four digesters at each time point relative to the previous sampled community. Operational stages correspond to the periods (■) Start-up, (■) Transition, and (■) Steady...... 101 Figure A-VI. % Relative Abundance (RA) explained by unique OTUs (Coloured), shared between 2 digesters (light grey), shared between 3 digesters (dark grey), and shared among the 4 digesters (black). In Bacterial kingdom brown shadowed area represents the %RA of archaeal populations. In Archaeal kingdom, pale-blue shadowed area represents the % RA belonging to bacteria. Operational stages correspond to the periods (■) Start-up, (■) Transition, and (■) Steady...... 102 Figure A-VII. Individual PCA ordinations for each digester and Procrustes analyses at OTU level (Hellinger transformed rarefied data). Each coloured circle represents one digester, and circle size increases with operational time. The axes and arrows indicate the translation and rotation of the PCA plots (pairwise comparison) ...... 103 Figure A-VIII. Heatmap 97% similarity. Non-filtered taxa for the four digesters at the 10 sampling points. Right column shows the lowest possible taxonomic assignment, and phylum level at left column. Darker intensity indicates higher relative abundance, grey cells indicate that this taxa was not detected in the samples at that level of resolution ...... 105

xvi Figure A-IX. Complete correlation maps taxa – metabolic activity rate. Left Pearson correlations, Right Spearman correlations...... 105 Figure A-X. Evolution of -diversity indexes with time for the four digesters. Day 0 represents the original inocula. (A) Estimated richness based on 16S OTUs clustered at 97% similarity. (B) Simpson index (0 represents complete dominance and 1 complete evenness). (C) Shannon-Wiener Entropy index…………………………………………………………………………………….107 Figure B-I. Individual VFA yields of the four digesters (♦) SL, (♦) SS, (♦) PL, (♦) BG at each SRT (■) 15-d and (■) 8-d SRT, (■) 4-d and (■) 2-d SRT. (A) Acetate and (B) Propionate. Data within dotted red lines has been excluded from the data analysis due to operational complications ...... 109 Figure B-II. Individual VFA yields of the four digesters (♦) SL, (♦) SS, (♦) PL, (♦) BG at each SRT (■) 15-d and (■) 8-d SRT, (■) 4-d and (■) 2-d SRT. (A) i-Butyrate, (B) n-Butyrate, (C) i-Valerate, (D) n-Valerate, and (E) Caproate. Data within dotted red lines has been excluded from the data analysis due to operational complications ...... 110 Figure B-III. Monitoring parameters of the four digesters (♦) SL, (♦) SS, (♦) PL, (♦) BG at each SRT (■) 15-d and (■) 8-d SRT, (■) 4-d and (■) 2-d SRT. (A) TS, (B) VS, (C) tCOD, (D) sCOD, and (E) pH. Data within dotted red lines has been excluded from the data analysis due to operational complications ...... 111 Figure B-IV. Evolution of the COD mas balance profiles of the four digesters (♦) SL, (♦) SS, (♦) PL, (♦) BG at each SRT (■) 15-d and (■) 8-d SRT, (■) 4-d and (■) 2-d SRT. Values close to 0 indicate that the theoretical COD removal calculated from the methane production equals to the experimental COD removal. Data within dotted red lines has been excluded from the data analysis due to operational complications...... 113

Figure B-V. Relative activity rates normalised by cellulose rate. (Left figures) The ratio km,i/km,cel >1 indicates cellulose hydrolysis is the limiting step. (Right figures) The ratio km,i/km,ac >1 indicates that the conversion of acetate to methane is slower than the acetogenesis. (A) Butyrate to cellulose, (B) butyrate to acetate, (C) ethanol to cellulose, (D) ethanol to acetate, (E) propionate to cellulose, (F) propionate to acetate, (G) acetate to cellulose, (H) Formate to cellulose, and (I) dominance of the methanogenic pathway. Solid coloured line represents ratio for each digester (–) SL, (–) SS, (–) PL, (–) BG. Shadowed coloured box represents the 95% confidence interval of the value ...... 115 Figure B-VI. Visualisation of PERMANOVA results testing differences among SRT (groups) on the microbial populations (OTUs Hellinger transformed) of the four digesters. Two possible clusters considering the centroid and dispersion were observed (■) 15-d and (■) 8-d SRT, (■) 4-d and (■) 2-d SRT...... 116 Figure B-VII. PCA of the microbial community profiles at the OTU level (Hellinger transformed) for the four digesters individually over 16 sampling events (4 at each SRT). Each sample is represented by a single circle, coloured according to digester designations in previous plots. Circle size decreases for each SRT. OTUs are presented as black crosses and the populations contributing most to the variability between microbial communities are identified. Each sample is represented by a single circle, coloured according to digester designations in previous plots (A) SL, (B) SS, (C) PL, and (D) BG...... 117 Figure B-VIII. PCA of the microbial community profiles at the OTU level (Hellinger transformed) for the four digesters over 16 sampling events (4 at each SRT). Each sample is represented by a single circle, coloured according to digester designations in previous plots. Circle size decreases for

xvii each SRT. OTUs are presented as black crosses and the populations contributing most to the variability between microbial communities are identified. Correlation with performance parameters are indicated by arrows, blue arrows represent those parameters correlated with a significance lower than 0.01, and orange arrows parameters with a significance lower than 0.05. (A) Bacterial community. (B) Archaeal community...... 118 Figure B-IX. Microbial composition at 97% similarity (relative abundance), observed OTUs, and diversity indexes (Shannon and Simpson index). Heatmap shows the populations in the four digesters at the 17 sampling points. The lowest possible taxonomic assignment is shown in the right column and phylum level in the left column. Microbial composition: Darker intensity indicates higher relative abundance, grey cells indicate that this taxa was not detected in the samples at the level of resolution (>0.005% of total counts at least in one sample). Observed OTUs: Darker intensity indicates higher observed OTUs per sample. Diversity indexes: Darker intensity indicates greater evenness (microbial populations more evenly distributed)...... 119 Figure B-X. Biofilm pictures taken during digesters dismantling (day 401, cycle 36) of the four digesters (A) SL, (B) SS, (C) PL, and (D) BG. Details of covering walls, paddle or aggregates that came of the paddle during the opening process...... 120 Figure B-XI. (A) First-order CSTR model fitting methane yield as function of SRT. Marker represent the average experimental methane yield at each SRT. Solid line represents the modelled profile. Each colour represents one digester (♦) SL, (♦) SS, (♦) PL and (♦) BG. (B) Comparison of first order kinetic constant (kh) obtained from the activity assays (green) vs the kinetic constant derived from the first-order CSTR model (black). Solid line represents the value of the kinetic constant, shadowed box represents the 95% confidence interval of the value ...... 121 Figure C-I. (Left) Cumulative specific methane production curves after pre-fermentation at different temperatures, treatment time, and control. (Right) Confidence regions from BMP test modelling after pre-fermentation and control. Each ellipse bounds the confidence region (95%) of each trial for the anaerobic biodegradability (f, x-axis) and the hydrolysis rate (khyd, y-axis) ...... 125

xviii List of Tables Table 1. Model substrates for activity assays ...... 29 Table 2. Characterisation of the two PS batches collected (average ± standard deviation) ...... 30 Table 3. List of materials and methods used in Chapter 5 previously described in Chapter 4 ...... 41 Table 4. Effluent characteristics and process efficiency for each digester at steady state conditions (average ± standard deviation) ...... 43 Table 5. List of materials and methods used in Chapter 6 previously described in Chapter 4 ...... 55 Table 6. List of materials and methods used in Chapter 7 previously described in Chapter 4 ...... 70 Table 7. Summary of solubilisation and acidification yields and methane yields for all the fermentation conditions ...... 74 Table A-I. Summary statistics from pairwise Procrustes analyses...... 104 Table B-I. Operational conditions and process performance for each digester at steady state conditions at each SRT (average ± standard deviation) ...... 112 Table C-I. COD balance for the combined fermentation and anaerobic digestion process ...... 126

xix List of abbreviations Ac Acetate mRNA Messenger ribonucleic acid

AD Anaerobic digestion N2 Nitrogen AM Acetoclastic methanogenesis Na+ Sodium cation ANOSIM Analysis of similarities NADH Nicotinamide adenine dinucleotide (reduced)

ANOVA Analysis of variance NaHCO3 Sodium bicarbonate + ARDRA Amplified ribosomal DNA restriction NH4 Ammonium cation analysis ATP Adenosine triphosphate NMDS Non-metric multidimensional scaling BG Anaerobic granules treating brewery OLR Organic loading rate wastewater BMP Biochemical methane potential test OTU Operational taxonomic unit Bu Butyrate PCA Principal component analysis Ca2+ Calcium cation PCR Polymerase chain reaction cDNA Complementary deoxyribonucleic PL Digested pig manure (lagoon) acid

CH4 Methane Pro Propionate

CO2 Carbon dioxide PS Primary sludge (municipal) CoA Coenzyme A PyNAST Python Nearest Alignment Space Termination COD Chemical oxygen demand QIIME Quantitative Insights Into Microbial Ecology CSTR Continuous stirred tank reactor RNA Ribonucleic acid DGGE Denaturing gradient gel rRNA Ribosomal ribonucleic acid electrophoresis DNA Deoxyribonucleic acid sCOD Soluble chemical oxygen demand EMP Embden-Meyerhof-Parnas (pathway) SL Digested slaughterhouse waste

Et Ethanol SRT Solid retention time FISH Fluorescence in situ hybridisation SS Digested sewage sludge For Formate T Temperature GH Glycoside hydrolases tCOD Total chemical oxygen demand

H2 Hydrogen T-RFLP Terminal restriction fragment length polymorphism HM Hydrogenotrophic methanogenesis TS Total solids HRT Hydraulic retention time VFA volatile fatty acids hyd Hydrolysis VS Volatile solids K+ Potassium cation WWTP Wastewater treatment plant

LCFA Long chain fatty acids

Chapter 1

1 Introduction

1.1 Project Significance Consumption patterns of modern society lead to the production of large and persistent amounts of biodegradable waste materials that need to be managed, treated, and disposed adequately to decrease the negative impacts on the environment and health (Hoornweg and Bhada-Tata 2012). These impacts include greenhouse gas emissions, groundwater contamination, the spread of pathogenic organisms and odours.

Biodegradable waste can be broadly defined as a waste stream that is capable of undergoing aerobic or anaerobic biological decomposition (European-Union 1999), where in urban societies these are mainly the organic fraction of municipal solid waste, commercial and industrial waste from food processors and packagers and sewage sludge generated from wastewater treatment.

Traditionally, the management of these wastes has been focused on the treatment-before-disposal to decrease the costs of associated with disposal (e.g. levies and transport costs) and alleviate the several adverse environmental effects. However, there is a growing trend to consider these biodegradable wastes as a potential source of renewable energy and as raw materials to produce valuable products that are the core of a circular bio-based economy (Puyol et al. 2016)

While many technologies are contributing to the challenge of resource recovery, anaerobic digestion (AD) is a widespread bioprocess to reduce waste volumes and produce methane rich biogas

1 that can be captured as a source of renewable energy (Mata-Alvarez et al. 2000). AD is a microbial- mediated, multi-step process where organic substrates are hydrolysed, fermented, and converted to biogas in the absence of oxygen. AD is a versatile microbial process and has been implemented in a large number of technologies to treat complex and diverse substrates across a broad range of industries. AD is widely used to stabilise sewage sludge in wastewater treatment plants, the organic fraction of municipal solid waste in mechanical-biological treatment plants, wastes from agriculture and livestock (crops, manure), or waste from food processing industries. However, with the raising

interest in the emerging bio-economy and circular economy strategy (OCDE 2009), Kalmykova et al., 2017) AD can be applied to recover intermediates of the digestion process have applications as commodity chemicals or as precursors to a range of bioproducts (e.g. biofuels, bioplastics, biotextiles, and biochemicals).

However, bioproducts are currently still expensive compared to fossil carbon equivalents (Kircher 2014). For this reason, there are considerable economic drivers for the optimisation of processes for bulk bio-commodities with a focus on maximising the utilisation of cheap and renewable resources. One strategy to achieve more cost-efficient bioprocesses and lower carbon footprints is to use biodegradable waste as a feedstock in bio-refineries. The bio-refinery concept embraces several technologies and bioprocesses that integrate feedstock conversion to chemicals, energy, biofuels, or other value-add products (Cherubini 2010, Murali et al. 2017). The carboxylate platform is an example of an attractive bioprocess to convert organic waste into short-chain fatty acids (traditionally known as volatile fatty acids) by anaerobic mixed culture fermentation (Agler et al. 2011, Cope et al. 2014). The carboxylate platform is an application of AD where the methanogenesis step is avoided. However, one of the main challenges for producing higher value organic compounds (as compared to biogas) from AD is the difficulty associated with robustly controlling mixed-culture fermentation products from a mix-culture microbial process.

1.2 Project rationale and methodological approach Current research on the microbial ecology of AD processes has advanced into the linkage between microbial community and digester performance, with the goal of bridging the gap between microbial ecology and mix-culture bioprocesses. However, the variety of AD configurations, operational conditions and substrate type; in addition to the complexity and dynamicity of associated microbial communities limit the understanding of the role that microbial populations play into the regulation of AD pathways that ultimately influence the process performance and product distribution.

2 An increased understanding of the microbial-process interaction can provide strategies to drive mixed-culture bioprocesses towards a desired commodity, and potentially reveal new avenues through process manipulation.

The focus of this research is to improve the understanding of the role of microbial populations in each step of the AD process. To gain this understanding, the results from laboratory scale anaerobic digesters, kinetic assays and culture-independent molecular methods were linked using multivariate statistical analyses.

Chapter 2

2 Literature Review

This chapter reviews literature that will develop the motivation of the research of this thesis. It starts presenting the background of anaerobic digestion from an engineering-driven perspective combined with an ecological-driven angle. The review is followed by the influence of several operational conditions on process performance and the molecular tools available to identify and track the dynamics of associated microbial communities and their functionalities.

2.1 Anaerobic Digestion Anaerobic digestion (AD) is a complex biochemical process where organic matter is degraded, in the absence of oxygen, into biogas (methane and carbon dioxide), digestate (stabilised organic matter), and liquor (soluble undegraded organics). AD requires the action of multiple trophic groups of microorganisms performing separate, but interdependent tasks in the overall degradation process. AD is often represented using four key biological steps: hydrolysis, acidogenesis, acetogenesis and methanogenesis (Figure 1) (Batstone et al. 2002).

Figure 1. Simplified scheme of the anaerobic digestion process. Adapted from Batstone et al. (2002)

From a microbiological point of view, AD can also be divided according to the functionality of its major microbial domains, Bacteria and Archaea. Bacteria perform the first steps of AD, hydrolysis, acidogenesis/fermentation, and acetogenesis, and they have the capability of metabolising a large variety of substrates under a wide range of operational conditions (Mata-Alvarez et al. 2014, Sundberg et al. 2013). Archaea, specifically methanogens, are responsible for the last metabolic step involved in the generation of methane.

2.1.1 Hydrolysis During hydrolysis, carbohydrates, proteins, and lipids are broken down into their monomers (i.e. sugars, amino acids, and long chain fatty acids) in a process catalysed by extracellular enzymes, traditionally grouped as cellulases, amylases, peptidases, and lipases depending on their function.

In the hydrolysis of complex carbohydrates, such as cellulose, a diverse range of enzymes are involved in the breakdown process (glycoside hydrolases, polysaccharide lyases, carbohydrate esterases), collectively designated, and classified within the Carbohydrate-Active enZymes database (CAZy) (Lombard et al. 2014). Among these enzymes, glycoside hydrolases (GH) are the most prominent group that cleaves the carbohydrates glycosidic linkages, which are expressed in a wide range of microorganisms, both aerobes and anaerobes. GH have been classified into families based on significant amino acid sequence similarity, where 145 GH families are found in the CAZy database

5 (AFMB et al. (last visit: 05/06/2017)). This classificationis a powerful predictive method for suggesting functions for newly sequenced enzymes encoded by a genome (Allgaier et al. 2010). In general, anaerobic bacteria perform hydrolysis in the vicinity of the particle, considered as a surface phenomenon (Sanders 2001). Some of the anaerobic cellulolytic bacteria have developed multi- enzyme complexes (cellulosomes) that depolymerise the cellulosic substrate and, at the same time, binds the bacterial cell wall to the particle surface, while others rely on cell-associated systems to adhere to the substrate particle (Figure 2).

A B

Figure 2. Simplified representation of cellulose hydrolysis mechanisms: (A) cellulosome system, (B) cell-associated system.

Cellulolytic activities are found in a large variety of anaerobic bacteria, for example the genera Clostridium, Acetivibrio, Butyrivibrio, Fibrobacter, Rumminicoccus, Thermotoga, Bacteroides, Bacillus, Myxoccocales (Lynd et al. 2002, Mullings and Parish 1984, Schnürer and Jarvis 2009, Vanwonterghem et al. 2014a) and recent genomic analyses have expanded our understanding to phyla Actinobacteria, Spirochaetes, Verrumicrobia, and Planctonomycetes (Gullert et al. 2016, Vanwonterghem et al. 2016) which usually coexist in biogas reactors (Lee et al. 2012, Sundberg et al. 2013, Vanwonterghem et al. 2014a).

The hydrolysis of proteins is carried out by enzymes termed peptidases, also known as proteases. Peptidases are substrate specific, so they show selectivity for the bonds they will hydrolyse, and can be grouped depending on: (i) the chemical mechanism of catalysis, (ii) the details of the reaction catalysed, or (iii) by molecular structure and homology; the latter one being the newest and most powerful classification (Barrett et al. 2013). Currently, there have been 367 peptidases families classified in the MEROPS database, based on the similarities in the primary (sequence of amino acids), and tertiary (three-dimensional structure) structure (WTSI (last visit: 05/06/2017)).

Research identifying proteolytic activity in microorganisms is less extensive than for cellulolytic, although, several anaerobic genera have been detected co-existing in AD communities with protein- rich feed. For example, Clostridium, Acetivibrio, Butyrivibrio, Thermotoga, Bacteroides,

6 Peptostreptococcus, Bifidobacterium, and Coprothermobacter (De Francisci et al. 2015, Palatsi et al. 2011).

In lipolysis, or lipid hydrolysis, extracellular lipases cleave the linkage between long chain fatty acids (LCFA) and the glycerol of the triacylglyceride (Palatsi et al. 2009). Most lipases act at a specific position on the glycerol bond of lipid substrate although they have the ability to catalyse a broad range of reactions (Arpigny and Jaeger 1999). Lipolytic activity in bacteria is found in the genera Bacillus, Pseudomonas, Staphylococcus, Dialister, Kyrpidia, or Syntrophomonas (Arpigny and Jaeger 1999, De Francisci et al. 2015, Palatsi et al. 2011). In a model mechanism for lipolysis, the lipases successively adsorb to the lipid surface, then they bind with the substrate, and next they perform the hydrolysis (Sanders 2001).

Finally, the hydrolysis products can be taken up by the initial degraders or other members of the community through cross-feeding (Flint et al. 2008, Müller 2008).

2.1.2 Acidogenesis / Primary Fermentation During acidogenesis, hydrolysis products are further converted into intermediate compounds such as volatile fatty acids (VFA), alcohols, ketones, formate or hydrogen. Sugars and amino acids are fermented while LCFA can only be anaerobically oxidised (mainly -oxidation) in a syntrophic relationship with hydrogen consumers. Fermentation is a redox process where the electron balance is maintained by the generation of fermentation products, and wasted electrons end up producing hydrogen (or formate) (Angelidaki and Batstone 2011, Madigan et al. 2010).

Fermentation of sugars and amino acids involves several metabolic pathways as well as diverse microbes that can yield several end-products. Many of these pathways have been studied in detail, for example, glycolysis via the Embden–Meyerhof pathway (EMP pathway). EMP glycolysis is the most relevant intracellular enzymatic pathway that breaks down glucose into the important precursor metabolites, pyruvate, energy (ATP), and reducing power (NADH) (Hoelzle et al. 2014, Madigan et al. 2010). Then, in mixed-culture fermentation, the most common subsequent pathway is the oxidation of pyruvate via pyruvate-derived acetyl-CoA branches, which have the capacity to generate ATP from substrate-level phosphorylation and regulate the internal redox processes through the production of a range of end-products. The most common end-products from metabolic acetyl-CoA branches are acetate, butyrate, ethanol, butanol, or acetone (Figure 3).

Figure 3. Simplified scheme of major metabolic pathways for carbohydrate fermentation. Orange envelope represents a bacterial cell. Adapted from Hoelzle et al. (2014), Madigan et al. (2010), Temudo et al. (2007)

Another important pyruvate-derived branch is the production of L-lactate. This branch is also widespread among microorganisms, which is the main pathway that leads to the production of lactate and propionate. Lactate can be excreted from the cells, or be further metabolised to propionate. (Figure 3).

Other alternative metabolic pathways are the Entner–Doudoroff pathway (a variant of EMP glycolysis) and the pentose phosphate pathway (Figure 3). Although they can share many intermediate structures, there are different enzymes involved as well as different energetic yields and reducing power that can shift the end-product distribution (Madigan et al. 2010).

A well-documented example of end-product regulation is the model organism Clostridium acetobutylicum. C. acetobutylicum is an acidogenic-solventogenic fermenter, known to regulate the endproduct spectrum depending on stress factors (e.g. pH change). As represented in Figure 4. at low pH the major products represent a mixture of acetone and butanol, while at pH closer to neutrality, the main products are acetate and butyrate.

Figure 4. Scheme of major metabolic pathways for glucose fermentation by C. acetobutylicum, depending on the environmental pH. Adapted from Dürre (2005).

However, in mixed-culture, the internal regulation mechanisms due to stress factors (i.e pH, temperature, inhibitors, H2 concentration) can be masked by the overall microbial community dynamics, since the growth and dominance of other species with different metabolic pathways can occur.

Amino acid fermentation commonly occurs through Stickland reactions where the oxidation of one amino acid is concomitant to the reduction of another amino acid. Despite the biochemistry behind the Stickland reaction being complex, from a metabolic strategy the reactions are simple (Figure 5). The oxidised amino acid is de-amined into pyruvate and ammonia, pyruvate is then catabolised to a fatty acid-CoA, and finally, a fatty acid is produced via substrate-level phosphorylation. The reduced amino acid is directly converted to the end-product via substrate-level phosphorylation (Madigan et al. 2010). The range of end-products depends on the type and concentration of amino acids present (Ramsay and Pullammanappallil 2001), which are acetate, propionate, butyrate (normal and branched), valerate (normal and branched), and organic aromatic acids.

Figure 5. Example of a Stickland reaction for the amino acid pair Alanine - Glycine. Orange envelope represents a bacterial cell. Adapted from Madigan et al. (2010)

Several genera of microbes are able to carry out fermentation pathways, for example, but not limited to, Clostridium, Lactobacillus, Lactococcus, Acetobacterium, Propionibacterium (Müller 2008). The fermentation end-products are excreted from the cells, which can serve as a substrate for acetogens and methanogens, or can be harvested from the process.

To summarise, during fermentation, individual microbes can utilise different metabolic pathways based on their enzymatic capabilities and, in at least some cases, they can shift between pathways to regulate internal electron and energetic flows based on stress factors. However, it is unclear how these factors impact to the ultimate end-product profile, especially when individual microbes reside in complex ecosystems.

2.1.3 Acetogenesis / Secondary Fermentation In acetogenesis, LCFA and intermediate compounds, other than acetate, are converted to acetate and hydrogen. Acetogenesis is carried out through -oxidation, where hydrogen (or formate) is both a reaction product and reaction inhibitor; consequently, hydrogen concentration must be limited to low levels for thermodynamic reaction feasibility. Therefore, this step is only possible through obligate syntrophy between acetogenic bacteria and H2-consumers (i.e. methanogens, sulphate reducing bacteria, and homoacetogens). Some of the known microbes that possess acetogenic activity belong to Clostridium, Syntrophomonas, Syntrophobacteraceae, Syntrophus, Acetobacteriu, and within Synergistetes (Drake et al. 2013, Mosbaek et al. 2016, Vanwonterghem et al. 2014a). Most

10 acetogenic species have been isolated from very diverse anoxic habitats and, if available, have the capability to metabolise other substrates, such as glucose. Moreover, these genera are closely related to other fermentative species, challenging molecular techniques to describe their function (Drake et al. 2013). Finally, reverse homoacetogenesis (where acetate is oxidised to H2 and CO2) despite being thermodynamically unfavourable, is well documented in anaerobic digesters, by syntrophic association with hydrogenotrophic methanogenesis; especially at extreme anaerobic environmental conditions (i.e. high ammonia or high temperatures) (Goberna et al. 2010, Ho et al. 2013, Karakashev et al. 2006).

2.1.4 Methanogenesis In methanogenesis, acetate, hydrogen (or formate) and carbon dioxide are converted to methane. Methanogenic archaea are a less diverse group than bacteria, because of their limited substrate range. Methanogens can be classified depending on the two main substrates used for methane production: (i) the strictly acetoclastic methanogens (Eq.1), which comprise the genus Methanosaeta, and (ii) the hydrogenotrophic methanogens (Eq.2) comprising the orders of Methanobacteriales, Methanomicrobiales, Methanococcales, Methanomassiliicoccales, and Methanopyrales. Moreover, the order Methanosarcinales, except the genus Methanosaeta, is considered a mixotrophic methanogen since can use either acetate, H2/CO2, methanol, or methylamines (Eq.3) to produce methane (Madigan et al. 2010, Mata-Alvarez et al. 2014).

0 -1 CH 3COOH  CH 4  CO 2 G = -31 kJ mol Eq.1 4H  CO CH  2H O G0 = -130 kJ mol-1 Eq.2 2 2 4 2 0 -1 CH 3 3 N  6H2O 9CH 4  3CO 2  4NH3 G = -76 kJ mol  Eq.3

11 2.2 Operational parameters The microbiology of AD involves a large phylogenetic diversity of bacteria and archaea that have different growth and rates, nutritional requirements, activities, and functionalities, which in balanced conditions decompose organic matter as previously illustrated in Figure 1. It is important to highlight that the balance of the different trophic groups will have a direct influence on process performance and intermediates accumulation. For instance, a negligible methanogenic activity will promote the accumulation of acetate and other VFA. Therefore it is important to determine how operational parameters and environmental conditions influence the dynamics of microbial groups and the formation of anaerobic products.

2.2.1 Inoculum The inoculum source has been studied from an operational point of view, since the quality and activity of the inoculum is crucial to avoid a process failure, especially during digester start-up (Regueiro et al. 2012).

Inoculum selection has primarily been assessed by determining the capabilities of the inoculum microbial community through short-term batch experiments (De Vrieze et al. 2015b, Gu et al. 2014, Langenheder et al. 2006, Perrotta et al. 2017). Although these tests provide an indicative response of the capabilities of the inoculum, they cannot predict if the characteristics will be maintained in the new environment or if the microbial community will be able to develop new capabilities.

Long-term batch experiments have shown that the adaptation of the inoculum to new substrates can improve the process kinetics, as well as the tolerance of microbes to inhibitors. Remarkable observations are found in Gavala and Lyberatos (2001), who acclimated the same inoculum to two different substrates, gelatine (protein) and lactose (disaccharide) over a long-term period. It was observed that sugar-acclimated inoculum degraded glucose faster although it preserved proteolytic function; the reverse was true for protein-acclimated inoculum. Moreover, the degradation of gelatine from the differently acclimated inocula yielded different proportions of acetate and propionate. Sugar- acclimated inoculum produced 4 mg of propionate (COD basis) per 1 mg of acetate (COD basis), while protein-acclimated inoculum produced 0.6 mg of propionate (COD basis) per 1 mg of acetate (COD basis). Since the assays were performed by limiting the amount of substrate in the spike, the metabolic response reflected the culture at that point in time, yet whether this activity would be maintained in long-term experiments remains uncertain.

In long-term continuous experiments, deterministic factors like substrate and operational conditions were demonstrated by (Liu et al., 2017, Lucas et al. 2015, Vanwonterghem et al. 2014a)

12 to drive anaerobic digester microbial populations. Over a period of more than 360 days, replicate digesters followed similar microbial dynamics and abundance correlated with changes in digesters performance. In a similar trend, Gomez-Romero et al. (2014) showed the drastic change in microbial population from the inoculum source (AD fed with fruit vegetable waste) when treating lactose as a sole carbon source. De Francisci et al. (2015) have recently studied the selective pressure of different substrates to modify the inoculum microbial community. Three digesters adapted to treat cattle manure at thermophilic conditions were exposed to three different model substrates, gelatine (protein), glucose (carbohydrate) and sodium oleate (lipid). When the digester was only fed with sodium oleate, a loss in microbial diversity was observed towards microbial specialisation, recognising the genus Dialister and Kyrpidia as key microbes to metabolise lipids. Moreover, a significant decrease in methanogenic activity occurred with the decline in methane yield. In contrast, glucose-fed digester showed a high microbial community dynamicity, where a deep microbial community transformation was observed, as well as an increase in community diversity. The co- occurrence and dominance of lactobacilli and Megasphaera elsdenii (able to transform lactate to propionate) were correlated with the accumulation of propionate in the glucose fed reactor. Finally, the bacterial community was resistant to the gelatine addition (non-relevant bacterial shift).

Although enrichments or adaptations to new conditions are documented in long-term studies, one of the remaining questions to solve is if differentiated initial communities (inocula) can converge due to deterministic factors to a similar community with comparable functionalities.

In this regard, Kim et al. (2013) traced the performance of two semi-continuous anaerobic digesters seeded with two different inocula and treating the same substrate (cheese whey). In their digesters, the number of bacterial copies in two reactors evolved in different ways while the digesters performance was comparable regarding VFA production, distribution, and biogas production. In addition, the quantitative interpretation of the Denaturing Gradient Gel Electrophoresis (DGGE) bands showed some variance between the two populations. Nevertheless, the dynamics of the number of bacterial copies cannot properly describe if the different inocula did not converged to a similar structure, but maintained the same functionality, and DGGE should be interpreted carefully, since quantitative data obtained from DGGE analysis can be biased from the generation of PCR-DGGE artifacts (Neilson et al. 2013). Even though no conclusive results can be extracted from that study, other questions can be formulated. Recently, Venkiteshwaran et al. (2017) used different AD inocula to treat synthetic wastewater at 3 g COD Lr-1 d-1 and 10 d SRT. Both studies reported the development of distinct microbial communities and bioreactor functionality in the replicates. It worth considering that in both studies inocula were subjected to the new conditions in a single step, which could have caused a severe disturbance to the microbial communities. Indeed, de Jonge et al. (2017) reported that

13 subjecting AD replicate digesters to a long starvation period resulted in different responses in community assembly and functionality once digesters operation was re-started. These contrasting results suggest that more systematic studies should be carried out to understand the undelaying mechanisms that drive the microbial assembly.

First, even if microbial communities are structurally dissimilar in different digesters under the same operational conditions; can the microbial functional redundancy maintain a similar digester performance? Second, is the initial inoculum a long-term variable for achieving a stable VFA production and distribution?

2.2.2 Substrate characteristics The substrate composition, the amount and structure of carbohydrates, proteins, and lipids, can shape the microbial populations in the digester. The microbial population may show different substrate affinity and preferential metabolic pathways to convert the substrate to very diverse end- products. This can be exemplified by the study carried out by Shen et al. (2014), who observed that the VFA product range, in batch anaerobic fermentation at 35 ºC, was dependant on the substrate composition (Figure 6). At the same time, the microbial populations differed from the inoculum at the end of the assay. This suggests that substrate is an influential factor to drive the development of the microbial community since the microbial relative abundance can change due to microbes’ substrate affinity. In continuous reactors experiments, Regueiro et al. (2014) and De Francisci et al. (2015) showed the influence of the substrate composition in the anaerobic digestion process. Briefly, both authors inoculated their digesters with the same inoculum source, and varied the feedstock characteristics. Regueiro et al. (2014) observed a high degree of clustering depending on the feedstock, where feedstocks that can support high active archaeal populations (i.e. soluble substrates) correlated with higher methane production. Similarly, the evaluation of microbial populations of full- scale digesters has found that digesters could be clustered together depending on the substrate characteristics (Smith et al. 2017, Sundberg et al. 2013, Werner et al. 2011b). The characteristics of the substrate also can affect the diversity, where carbohydrate-rich substrates offer a more diverse community than protein or lipid-rich substrates (De Francisci et al., 2015), probably due to a higher degree of specialisation of the associated communities. However, while microbial communities can adapt to a specific substrate, it is unclear if the diverse functional capacity is preserved or is adversely affected. In terms of functional capacity, Poszytek et al, (2017) showed the acclimation of different microbiomes to treat maize silage during the anaerobic digestion process. It was found that the microbial communities rapidly adapted to the substrate, being dominated by the representatives of Lactobacillaceae, Prevotellaceae and Veillonellaceae. After the adaptation period, the microbial

14 communities developed a higher hydrolytic capacity compared to the initial stages suggesting that substrate characteristics can favour the development or improvement of microbial functionalities.

Figure 6. VFA production and VFA distribution depending on substrate composition - 10 g CODinitial at 10 days. Adapted from Shen et al. (2014).

In addition, under fermentative conditions it has been observed that the same substrate can be degraded to several different endproducts (Figure 3) under different operational parameters (Horiuchi et al. 2002, Lu et al. 2011, Temudo et al. 2007), where another process such as temperature, pH, or solid retention time, may constraint either the microbial structure or the metabolic pathways. The control of these is still controversial in mixed culture processes, and the change in pathways is poorly understood in methanogenic systems (Junicke et al. 2016).

2.2.3 Temperature Temperature is one of the most important parameters that can affect anaerobic digestion processes, for either energy recovery or production of high-value compounds. Temperature can be related to the reaction and microbial growth rates, both increasing with temperature according to Arrhenius-type equation (Eq. 4); gas and salts solubility, changes in chemical equilibriums, and mass transfer rates. Moreover, the temperature is a parameter easily controlled at lab-scale as well as in full-scale plants.

 E   A  Eq. 4 k  Ae RT  AD can be effectively operated at a broad range of temperature including the following ranges, psychrophilic (> 20 ºC), mesophilic (25-40 ºC) and thermophilic (45-70 ºC). These ranges match with the microbiological temperature growth dependence, where microbes can withstand a range of 25- 40 ºC between the minimum temperature below which growth is not possible, and maximum temperature above which cell may be irreversibly damaged (death) (Madigan et al. 2010). Equally

15 important to the minimum and maximum temperatures, is the optimal temperature where the growth and enzymatic reactions occur fastest (Figure 7A). In that sense, AD focused on biogas production has been carried out at mesophilic conditions, operating at temperatures of 35-37 ºC since it is the optimal mesophilic temperature for methanogenic populations (Van Lier et al. 1997) and methane production rates (Fig 7B). Likewise, the optimum temperature at thermophilic conditions is around the 52-56 ºC (Ahring 1994).

A B

Figure 7. (A) Temperatures (minimum, maximum, and optimal) for microbial growth rate. (Adapted from (Madigan et al. 2010)). (B) Reaction rates of methane production from waste activated sludge at mesophilic range (Adapted from (Donoso-Bravo et al. 2009))

Several studies have evaluated and compared the performance of AD under mesophilic and thermophilic conditions aiming to optimise AD performance (Bolzonella et al. 2012, Bouskova et al. 2005, Kim et al. 2002, Nges and Liu 2009, Song et al. 2004). Others have evaluated the process resistance to temperature shocks (Ahn and Forster 2002, Peck et al. 1985, Speece and Kem 1970, Visser et al. 1993) even temperature fluctuations (Bourque et al. 2008, El-Mashad et al. 2004, Peces et al. 2013), generally from an overall process perspective. However, knowledge on the influence of temperature on different trophic groups and AD steps, and how this interacts and links with the overall process is scarce. Gavala et al. (2003) studied the impact of temperature (37 or 55 ºC) on the AD of primary sludge. While the acetogenic activity and methanogenic activity were the same order of magnitude at both temperatures, the accumulation of intermediates (acetate and propionate) were higher at 55 ºC compared to 37 ºC. Therefore, it can be hypothesised that hydrolysis and acidogenesis were faster than the uptake of acidogenic. It is agreed that hydrolysis is a temperature dependent step, increasing with temperature (Donoso-Bravo et al. 2009, Ge et al. 2011a, Siegrist et al. 2002). However, the temperature dependence of acidogenesis is less clear. Ge et al. (2011a) evaluated the relative kinetics of the four AD steps at 38, 55, 60, 65 and 70 ºC. Propionate acetogenesis and acetoclastic methanogenesis were not affected by temperature, in agreement with the activities calculated by Gavala et al. (2003), both using batch activity tests. However, glucose consumption

16 was faster at 38 ºC than at thermophilic temperatures, although kinetic parameters could not be estimated. This latter finding seems in contradiction with the hypothesis to explain the accumulation of VFA at thermophilic temperatures found in some of the literature comparative studies (Bouskova et al. 2005, Gavala et al. 2003, Kim et al. 2002, Sans et al. 1995). In this regard, more research is needed to study the kinetics of the main trophic groups concurrently with continuous digester operation. These studies can help to understand how the balance of AD steps are linked to digester performance for a given operational conditions.

Temperature not only affects the kinetics but it is also a selective pressure parameter to shape the microbial communities (Bialek et al. 2012, Pervin et al. 2013, Smith et al. 2017). Smith et al. (2017) found that full-scale thermophilic digesters had lower performance yields compared to mesophilic digesters, associated with the lower abundance of acetogenic bacteria and methanogens at 55 ºC. Pervin et al. (2013) examined the effect of temperature on a hydrolytic anaerobic digester operated at 35, 50, 60 and 65 ºC treating activated sludge. By characterising the microbial communities, it was shown that at 35 ºC the changes in the microbial composition corresponded to the changes in feed sludge microbial structure. On the other hand, at thermophilic conditions, the variation in microbial community was mainly due to temperature changes. For instance, at 50 ºC the consortium was dominated by Thermotogae, while at 65 ºC Thermotogae phylum was outcompeted by Lutispora thermophila and Coprothermobacter. The two latter microbes have a preference for fermenting + proteins, which correlated with the higher levels of NH4 in the digester medium. Bialek et al. (2012) monitored the microbial community profiles of a biogas digester treating synthetic dairy wastewater (i.e. no external microbial influx) at 37, 25, and 15 ºC. Once the inoculum was adapted to the new substrate at 37 ºC and steady operation was achieved, temperature was decreased to 25 ºC, and further on to 15 ºC. It was observed that microbial profiles were extremely close at 37 and 25 ºC as well as the digester performance while decreasing to 15 ºC promoted the dominance of homoacetogens. In a subsequent study performed by the same authors (Bialek et al. 2014) the temperature was decreased to 10 ºC, where the minor changes on the microbial community were attributed to the temperature decrease.

In addition to influencing the microbial structure, temperature is known to favour some metabolic pathways and syntrophic relations. Ho et al. (2014) demonstrated that syntrophic acetate oxidation coupled with hydrogenotrophic methanogenesis was the most important pathway to methane generation in extreme thermophilic temperatures. Acetate oxidation is endorgenic (G > 0) at standard conditions, hence is only possible if low levels of hydrogen are maintained. Hydrogen partial pressure is not the only factor influencing the thermodynamics, so is the temperature. For this specific pathway, the higher temperatures enhanced the thermodynamics free energy favouring the syntrophy

17 between acetate oxidation and hydrogenotrophic methanogenesis. Similar constraints could also happen at different levels microbial levels, favouring syntrophies, or cooperation, between microbes.

2.2.4 Retention time and organic loading rate Retention time can be defined as the average time that material remains inside the digester. Commonly, it can be divided in solid retention time (SRT) and hydraulic retention time (HRT). SRT stands for the average time that particles (solid substrates and microbial population) remain in the digester, while HRT is related to the digester hydraulic flow (liquid or soluble substrates). In well- mixed processes, such as continuous stirred tank reactors (CSTR), the SRT is equal to the HRT. SRT is usually defined as a design parameter and, its variation can impose some selective pressure over the system. Operating at short SRT may washout slow-growing microbes with doubling times higher than the operational SRT; however, a long SRT can reduce or remove this selected pressure. Nevertheless, a minimal SRT must be taken into account not only from a microbial perspective but also an operational imperative, since at a very low SRT it may not be possible to hydrolyse the substrate.

The organic loading rate (OLR) as a parameter is used to describe the rate at which substrate is added to the process. For high OLR, especially with easily biodegradable substrates, it may lead to a system overload and eventual process failure. Under an overloading period, the balance among the different microbial groups can be destabilised causing the accumulation of undesired intermediate products thereby affecting the substrate availability. Similarly to the SRT, the OLR has been defined as a process design parameter since both parameters are related (Eq.5), where S0 is the substrate concentration. Substrate concentration is usually inherent to the substrate characteristics, and its manipulation from a practical point of view is questionable since it implies the use of water (for dilution) or extra operational units (for concentration). However, the use of uncoupled SRT and OLR as regulatory parameters is still unexploited (Hoelzle et al. 2014). Recent studies have proposed the use of the OLR as a regulatory parameter to manage the microbial populations, were OLR shocks have shown to increase the microbial community resistance under stress conditions (Ferguson et al. 2016, Regueiro et al. 2015).

S0 OLR  Eq.5 SRT

Among the studies devoted to the changes in SRT/OLR, Krakat et al. (2011) did not observe substantial changes in product formation that could be correlated with changes in SRT between 37 and 8 days. However, these results could have been biased by (i) the simultaneous change in SRT and

18 OLR, (ii) the use of different substrates that could have masked the actual effect of SRT/OLR, (iii) the detection limit for microbial populations using amplified ribosomal DNA restriction analysis (ARDRA) was 105 cells mL-1, which could not have detected changes in minor populations. Nevertheless, an interesting finding was the emerging of Acidobacteria at the lower SRT (8 days) where the highest VFA concentration and the lowest methane production were recorded. Although methanogenic populations were not monitored, lower SRT could have started to restrict methanogenic function. Decreasing SRT has been proven to be effective in driving microbial communities towards the production of high levels of volatile fatty acids, showing a drastic accumulation when SRT was shortened to 4 days at both mesophilic and thermophilic conditions by restricting methanogenic populations (Vanwonterghem et al. 2015).

However, when already operating at acidogenic conditions, changes in VFA production may be limited by hydrolysis, since very low SRT can constrain hydrolytic activity. Maspolim et al. (2014), did not find severe changes in their mesophilic acidogenic reactor by changing SRT in the studied range (2 – 5 days) as shown in Figure 8A, where lower VFA yields at 2-d SRT could be attributed to lower solubilisation. The phylogenetic identification (Figure 8B) showed a different bacterial population compared to the seed sludge but similar within the three SRT that were studied, suggesting that SRT can drive microbial populations to a certain extent. Moreover, some of the microbes found in the reactor were also found in the substrate results that are in agreement with the ones found by Pervin et al. (2013).

A B

Figure 8. (A) VFA production and distribution and (B) DGGE gel band results for a mesophilic acidogenic digester running at 5, 3, and 2 SRT. Adapted from Maspolim et al. (2014)

19 2.2.5 pH pH will modify microbial growth and chemical equilibriums that may have an important effect in the overall anaerobic process. In general, a pH range around 6 - 6.5 are found as an optimum for hydrolysis/acidogenesis (Dareioti et al. 2014, Jagadabhi et al. 2010, Kim et al. 2003), while higher pH, around 7, is an optimum for methanogenesis (Appels et al. 2008, Whitman et al. 2006). Although most of the microbes can only live around neutrality (5.5 – 8), some microbes can grow in extreme conditions, acidophiles (< 5.5) and alkaliphiles (> 8) (Madigan et al. 2010).

The shift from butyrate to ethanol as a fermentation product has been documented to happen in mixed cultures fermenting glucose. Low pH was found to favour butyrate production while ethanol was the main product at alkaline pH. This trend using glucose as a sole carbon source has been observed by different authors and temporal scale (Lu et al. 2011, Temudo et al. 2007, Zoetemeyer et al. 1982), suggesting that the pH factor can have a stronger effect on regulating the metabolic pathways than actual different microbes. However, these results contrast with the findings by Horiuchi et al. (2002) and Fang and Liu (2002), where propionate was the favoured product at higher pH (7 and 8 respectively), while low and steady ethanol concentration was recorded over the entire pH range. Therefore, other operational parameters, such as temperature or retention time may have played an important role. In this regard, (Lu et al. 2011, Temudo et al. 2007, Zoetemeyer et al. 1982) operated the reactors at 30 ºC, while (Fang and Liu 2002, Horiuchi et al. 2002) at 36-37 ºC. These findings highlight the importance of parameters interactions and the need for systematic studies where conditions can be replicated.

However, in AD systems pH is stable due to the inherent buffering capacity of the process. The major drawback of pH manipulation is that although easy to control in laboratory scale by adding acid or alkali, the full-scale implementation may be hindered by the extra costs of pH control chemicals. Moreover, in systems with strong buffering capacity (mainly due to bicarbonates, and ammonia equilibriums), high amounts of chemicals might be used, which besides the economic costs, the extra metal ions added through alkali (i.e. Na+, K+, Ca2+) may inhibit the anaerobic process if certain thresholds are surpassed.

2.3 Linking microbes and process: Molecular tools and statistical methods As AD is a microbiologically mediated process, there has been a special interest for decades to study the interactions between microbial populations and reactor performance. Early studies relied on culture-dependent techniques, which despite the valuable identification of specific key microbes and metabolic processes (Carballa et al. 2015, Vanwonterghem et al. 2014b), held biased results due

20 to the selective cultivation, isolation, and enrichment of microbes from their natural habitats (Marzorati et al. 2008). The need for identifying and monitoring microbes in their habitat launched the development of culture-independent techniques. Early AD microbial-driven studies have been mostly tackled from a descriptive perspective, showing ‘who is in there’ but the need to further improve the understanding of the microbial communities has inclined research towards the adoption of numerical ecology methods (multivariate analyses) to explore trends, correlate data, and ultimately test and draw new hypothesis (Legendre and Legendre 2012).

2.3.1 Phylogenetic characterisation Fluorescence in situ hybridisation (FISH) allows the in-situ identification and visualisation of microbes by targeting the 16S ribosomal RNA (rRNA) (Hugenholtz et al. 2002, Wagner et al. 1993). FISH allows the direct examination of key microorganisms and their spatial distribution; however, it is dependent on the 16S rRNA probe. This method, therefore, cannot identify unknown taxa. FISH probes can be designed to target diverse microbial groups (e.g. all bacteria) or very specific microbial groups (e.g. at the operational taxonomic unit (OTU level) and any level of specificity in between. FISH probes can identify the abundance and spatial location of microbes covered by the selected probe but will fail in providing any information on microbes not covered by the probe. Therefore, this limits the usefulness of FISH probes in diverse and largely unknown microbial communities.

Other culture-independent methods that target the rRNA, known today as traditional molecular techniques (e.g., DGGE, terminal restriction fragment length polymorphism (T-RFLP), ARDRA). These fingerprint techniques offer semi-quantitative insights of community composition, especially useful for monitoring the dynamics of microbial communities over time. Nevertheless, they are limited in their ability to provide taxonomic identities. If using these techniques, further analysis is required to assign fingerprint outputs to a taxonomical unit. The construction of clone libraries based on 16S cloning and Sanger sequencing provides the phylogenetic evaluation. However, its application is laborious and costly, with the result that outputs are limited to about 100 sequences per sample in most applications, typically concentrating on dominant populations (Cirne et al. 2007, Degnan and Ochman 2012, Regueiro et al. 2012). The relatively small number of clones makes it very difficult to evaluate minor or rare taxa.

More recently, the development of next-generation sequencing techniques, based on high- throughput sequencing platforms (e.g. Roche 454 or Illumina©) have allowed advancements including: (i) increased data to thousands of sequences per sample and a substantially higher microbial resolutions (ii) processing larger amount of samples, (iii) decreased the processing time, and (iv) lower sequencing costs (Hugenholtz and Tyson 2008, Vanwonterghem et al. 2014b). The

21 sequencing of 16S rRNA gene amplicons by high-throughput sequencing platforms have revealed a greater microbial diversity that so far has been undetected due to methodological constraints, revealing even rare taxa and further improving the understanding on how the community is driven by the abiotic environment (Pinto and Raskin 2012, Sundberg et al. 2013, Werner et al. 2011a).

However, a common bias for all these techniques is related to the use of polymerase chain reaction (PCR) to carry out the gene amplification. Despite constant advances to overcome these disadvantages some biases can come from: (i) sequencing errors and chimeric PCR artifacts that inflate the estimated richness of the sample, (ii) a large portion of the microbial diversity in a sampled community may not be captured due to primer mismatches, and (iii) differential amplification efficiencies of the 16S rRNA genes in multi-template PCR reactions that can influence the relative abundances of detected OTUs (Pinto and Raskin 2012).

Moreover, the 16S rRNA sequencing techniques cannot sequence the entire genome and cannot be directly used to infer metabolic functionality. New approaches have been developed to sequence the entire genome pool directly from a natural or engineered environment known as meta-omics (Hugenholtz and Tyson 2008, Morgan and Huttenhower 2014, Vanwonterghem et al. 2014b), to provide a deeper understanding of fundamental biochemical processes.

Metagenomics, is a DNA-focused approach that provides a deeper description of the phylogenetic assemblage, overcoming the biases from PCR primers and can be used to predict microbial functions. The ultimate goal of metagenomics is to reconstruct large genome fragments from highly complex environments, including anaerobic digesters, aim to determine how individual populations interact and contribute to the whole consortium (Vanwonterghem et al. 2014b).

2.3.2 Functional capacity and expression: Advanced molecular techniques While metagenomics can be used to predict potential functions, it is limited to the previous known function deposited in the database (Handelsman 2007), sequencing technologies and bioinformatics pipelines (Forbes et al., 2017). In addition, genes regulating a certain function may not be expressed at any given point in time. To unravel the microbial functional capabilities, metatranscriptome analysis focuses on the messenger RNA (mRNA), which provides a method to measure a snapshot of the in situ gene expression. Metatranscriptomic outputs need to be integrated with metagenomic analyses from the same environment to ascertain that changes in function that are the result of changes in taxa and to discover highly expressed pathways (Morgan and Huttenhower 2014, Vanwonterghem et al. 2014b). Moreover, metatranscriptomics can be used to understand why and how a microbial community shifts its metabolic pathways as a result of changes in environmental conditions.

22 However, with the more complex and detailed data recovered from these methods, arise a greater difficulty in analysing and interpreting the data obtained (Morgan and Huttenhower 2014). To unravel the complexity of sequencing outputs from this method, researchers in bioinformatics have developed software pipelines (e.g., QIIME, GroopM) to facilitate the data assemblage and interpretation (Caporaso et al. 2010b, Imelfort et al. 2014). Even with these software pipelines, these techniques are closely dependent on the efficiency and quality of the DNA or RNA recovered (Stewart 2013).

Other meta-omic approaches are found to provide additional levels of microbial regulation. Metaproteomics, based on identifying peptide fingerprints throughout the protein extraction, can be used to detect catalytic enzymes, entire metabolic pathways, and novel functional proteins. Metabolomic analyses involve the monitoring of all molecules produced by cells, which can be involved in cell anabolism or catabolism. Metabolomics can give a comprehensive view of how cells function, as well as identifying changes in specific metabolic pathways. However, the bottlenecks of the application of meta-omics are related to the difficulties to extract and measure a wide range of proteins and metabolites, and conservation of mRNA (Morgan and Huttenhower 2014).

Other sequencing technologies, such as Nanopore DNA sequencing Jain et al. (2016) was considered established in 2016, where a portable device can sequence genomic DNA, cDNA and RNA in real time with high reading lengths with an accuracy >92%, and expanding to protein detection. The relatively low cost of this technology together with the benefits of direct sequencing (without PCR amplification), deeper sequencing, and real-time data; makes it a promising technology to resolve questions pertaining to the challenging field of characterising the functional capacity of an ecosystem in the upcoming future.

2.3.3 Application of high-throughput technologies in the anaerobic digestion process The decreasing costs of high-throughput technologies, the increasing computational power, and the development of methods to improve the quality of DNA, RNA, proteins and metabolites recovered are rapidly improving the understanding of the microbial communities in many environments (Cardenas et al. 2016, Widder et al. 2016, Zampieri et al. 2017).

Targeting the anaerobic digestion process, Zakrzewski et al. (2012) and Hanreich et al. (2013) pioneered the application of meta-omic approaches to unravel the mechanisms of the biogas formation from agri-industial waste. Hanreich et al. (2013) indicated that in their batch assays the hydrolysis of cellulosic compounds was mainly carried by Firmicutes phylum as shown by their high expression of glycoside hydrolases, while Bacteroidetes showed a high number of sugar transporters suggesting a key involvement in the degradation of other polysaccharides. Zakrzewski et al. (2012) found a high expression of transcripts for methanogenic enzymes, suggesting to the active

23 methanogenic pathway during the steady phase of a full-scale digester. Interestingly, the most abundant species dominating the community also contributed the majority of the transcripts, suggesting a correlation between abundance and activity in the studied system. The rapid improvements in bioinformatics software to process the data from metagenomic surveys, has improved the understanding the ecology of the anaerobic digestion microbiome. Metagenomic studies have been applied to reconstruct genomes from uncultured microorganisms abundant in some anaerobic digestion bioreactors. Kirkegaard et al. (2016) found that the candidate phylum Hyd24-12 (named Candidatus Fermentibacteria), abundant in mesophilic anaerobic digesters treating municipal wastewater sludge, is likely to be involved in the acidogenesis step; producing acetate and hydrogen from the fermentation of sugars. Moreover, it is able to utilise sulphate as an electron acceptor, suggesting that it might be involved in the formation of hydrogen sulphide during the AD process. Using a similar approach, Maus et al., (2016) described Defluviitoga tunisiensis (phylum Thermotogae) as an abundant player in the utilisation of carbohydrates and fermentation of sugars in the thermophilic AD. The reconstruction of fermentation pathways from Clostridium bornimense predicted that during the degradation of maize silage, C. Bornimense had the potential to ferment xylan and xyloglucan to a variety of fermentation products such as hydrogen, acetate, formate, lactate, butyrate, and ethanol (Tomazetto et al., 2018). The combination of metagenomics with other techniques (i.e. isotope fractionation) has revealed new syntrophic-acetate oxidation organisms under the accumulation of acetate. For instance, Mosbaek et al., 2016 found that the consumption was facilitated by the activity of Methanosarcina and Methanoculleus and five subspecies of Clostridia, the latter containing the gene formyltetrahydrofolate synthetase, a key enzyme for reductive acetogenesis. Similarly, Ruiz-Sanchez et al. (2018) found that syntrophic acetate oxidation pathways could be accomplished by some representative of the phyla Chloroflexi and Bacteroidales via the glycine cleavage system, to date undescribed to perform the syntrophic acetate oxidation pathway.

The combination of metagenomics with targeted FISH has revealed that the uncultured taxon A6 (within the phylum Chloroflexi) has a fermentative metabolism and it is located alongside with Methanosaeta spp. suggesting potential synergistic relationship (McIrloy et al., 2017). Nonetheless, all these findings aim to describe the metabolic routes of a single organisms or syntrophic partnerships. Several authors have used metagenomics from a whole process perspective, describing the main metabolic pathways responsible for the AD process and the main intermediates production linked to their concomitant microbial populations (Vanworterghem et al., 2016, Treu et al., 2016, Treu et al., 2016b, Campanaro et al., 2017, Fontana et al., 2018). Although the new insights provided by the studies using high-throughput techniques, a great number of the studies here reviewed lack of information related to the AD process operation and performance. This information is particularly

24 useful for cross-comparison between studies, since as shown in section 2.2, operational parameters can shape AD functionality.

2.3.4 Microbial to process. Integrated approaches and statistical methods The improvement and development of microbial molecular tools have undoubtedly expanded our knowledge on the structure of microbial communities in complex ecosystems. However, the linkage to process functionality requires the combination of different disciplines and techniques to reliably correlate ‘who does what’ to overcome limitations of each technique.

Figure 9. Combination of techniques to gain an understanding of bioprocesses

For instance, the combination of molecular-imaging techniques (i.e. FISH) with the addition of labelled isotopes (MAR-FISH, NanoSIMS) has provided insights into which organisms are active under certain operational conditions, and syntrophic associations (Ho et al. 2014). Although these approaches can provide a link between microbial populations and specific metabolic process, they cannot measure relative microbial activities, such as the kinetics of substrate uptake, product formation, or yields. In this sense, the combination of process monitoring data (product yields and kinetic analyses) with molecular techniques open alternatives to disentangle ‘who and how is doing what’. Moreover, bridging the gap between microbial ecology and process engineering not only requires the combination of experimental techniques but also requires the application of qualitative and quantitative approaches to answer the questions and hypothesis aroused. To comprehensively approach these questions, microbial ecologists rely on multivariate statistical analyses to overcome the obstacles of large data sets, visualise and summarise trends and further explore microbial relationships (Legendre and Legendre 2012, Ramette 2007). For example, these numerical tools aid the exploration of the distribution and diversity patterns (dynamics) of microbial communities (e.g., Principal Component Analysis, Clustering Analyses), and establish hypothesis-driven approaches (e.g. constrained ordination analysis, Mantel, ANOSIM) that test for the relationships between variables.

Chapter 3

3 Research gaps and objectives

3.1 Research gaps From the review of existing literature, there is a knowledge gap in fundamentally understanding the influence of initial microbial consortia on long-term process operation, stability and product profile. Although enrichments or adaptations to new conditions are documented in long-term studies, one of the remaining questions to solve is if differentiated initial communities (inocula) will converge due to deterministic factors, resulting in similar communities with comparable functionalities. In addition, it remains uncertain how the microbial functions are preserved or adversely affected in response to process changes, and how this is displayed by different process yields or end-product recovery. Such changes could influence the process to different degrees, by changing the microbial populations and their metabolic pathways, decreasing the microbial richness but maintaining similar functions, or losing complete functions that could drive the process to the desired point (e.g., an acidogenic phase as opposed to a methanogenic phase) or even cause the failure of the overall process. Therefore, understanding how the change in microbial populations or loss in microbial functionality can influence the behaviour of the rest of the community can prove to be a powerful tool for manipulating and controlling processes towards a desired commodity.

26 3.2 Research objectives The research objectives presented in this thesis have been designed accordingly to address these gaps with the main goal of understanding the changes in microbial communities linked to the anaerobic digestion process, from a performance and kinetic perspective. Each research objective is developed in one chapter of this thesis:

Research objective 1 (Chapter 5): To determine how the seeding microbial community (inoculum) and its characteristics will develop through anaerobic digestion long-term operation under balanced conditions, and whether this is reflected in process performance and process kinetics.

Research objective 2 (Chapter 6): After the microbial communities are fully acclimatised to the AD operational conditions (Research objective 1), to determine the influence of solid retention time as a sole selection pressure to drive the development of the microbial communities, and its effect on process performance in terms of product yield, end-product distribution, and activity (metabolic) rates.

Research objective 3 (Chapter 7): Implementation of operational conditions to drive the formation of volatile fatty acids and subsequent methanisation of the remaining solid fraction in an integrated bio-refinery concept.

Chapter 4

4 Materials and methods

This chapter describes the experimental set-up, materials and methods to achieve the research objectives presented in Chapter 3. The research presented in this thesis has been conducted on two experimental set ups. The first experimental set up consisted on four parallel digesters, each one inoculated with a different inoculum source, which were used to monitor AD process performance (yields and stability) and serve as inoculum to determine the kinetics of the main metabolic steps in separate batch activity assays (Figure 10). The second experimental set up consisted on batch fermentation assays followed by biochemical methane potential (BMP) tests to implement operational driven decisions (Figure 11).

4.1 Inoculum sources and substrates

4.1.1 Inocula Inocula were collected from the following sources: (i) SS: Mesophilic anaerobic digested sewage sludge from a conventional municipal wastewater treatment plant (Queensland, Australia). The SS was collected from a digester that treats mixed sewage sludge (50% primary and 50% secondary sludge on VS-basis) at a solid retention time of 23–24 days and under mesophilic conditions (T=35±2 °C).

(ii) PL: Inoculum was collected from the sludge layer (∼2 m) of a semi-covered piggery lagoon (PL) (Queensland, Australia) that receives flushed, unscreened manure from a specialised breeder piggery. The lagoon is operated at high retention times and ambient temperature.

28 (iii) SL: Digestate from a crusted anaerobic lagoon at a cattle-only slaughterhouse (New South Wales, Australia) that receives combined wastewater (slaughter floor, boning room, cattle wash, rendering, paunch) after primary treatment to remove coarse solids and reduce the fat content. The lagoon is operated at low hydraulic retention times and has an uncontrolled mesophilic temperature profile (38±2 °C).

(iv) BG: Anaerobic granules from an upflow anaerobic sludge blanket (UASB) reactor at a brewery (Queensland, Australia) that treats acidified brewery wastewater under mesophilic conditions at an HRT of 0.5 days.

4.1.2 Substrates 4.1.2.1 Digesters substrates and media The four digesters were fed synthetic media containing a mixture of microcrystalline cellulose (Sigma-Aldrich, USA) and casein from bovine milk (Sigma-Aldrich, NZ) at a ratio of 75:25 (COD- basis), anaerobic basic medium Angelidaki et al. (2009) and supplemented with sodium bicarbonate -1 -1 buffer solution (0.35 g NaHCO3 Lr d ).

4.1.2.2 Activity assays substrates Six model substrates were chosen to represent three main AD metabolic steps (Table 1). Microcrystalline cellulose (Sigma-Aldrich, USA) as representative of hydrolysis. Sodium butyrate (Sigma-Aldrich, IN), sodium propionate (Sigma-Aldrich, USA) and Ethanol Absolute (Merck, Germany) were used as representatives of acetogenesis. Sodium acetate (Sigma-Aldrich, USA) and sodium formate (Sigma-Aldrich, USA) were used as representative of acetoclastic methanogenesis (AM) and hydrogenotrophic methanogenesis (HM), respectively.

Table 1. Model substrates for activity assays Substrate Targeted metabolic activity Cellulose Hydrolysis Butyrate Acetogenesis Ethanol Acetogenesis Propionate Acetogenesis Acetate Acetoclastic methanogenesis Formate Hydrogenotrophic methanogenesis

29 4.1.2.3 Fermentation assays substrates Primary sludge (PS) was obtained from the same municipal wastewater treatment plant that inoculum SS (Queensland, Australia). PS was collected after being thickened by centrifugation and before being mixed with waste activated sludge and fed into the mesophilic AD treatment. PS was used for fermentation batch experiments immediately after collection. The solid fraction of the fermented sludge was stored at 4 ºC before biochemical methane potential (BMP) testing (max. 4 days). The second batch of PS was collected from the same municipal wastewater treatment plant replicate and validate the results obtained from the first fermentation trial. Table 2 summarises the main characteristics of the two sets of PS used in this study.

Table 2. Characterisation of the two PS batches collected (average ± standard deviation)* Parameter Units Batch 1 Batch 2 TS gTS L-1 47.9 ± 0.3 52.9 ± 0.2 VS gVS L-1 41.6 ± 0.2 45.4 ± 0.2 tCOD gCOD L-1 57.9 ± 0.6 72.1 ± 2.7 sCOD gCOD L-1 2.0 ± 0.1 2.4 ± 0.1 pH - 5.1 ± 0.1 4.9 ± 0.1 VFA gCOD L-1 2.0 ± 0.1 2.4 ± 0.1 Alcohols gCOD L-1 b.d.l** b.d.l COD:VS - 1.39 1.59 * n = 4 (number of replicates to calculate the average) ** b.d.l., below detection limit (< 1 mg L-1)

4.2 Experimental Set up 4.2.1 Continuous digesters

Four digesters, each with a working volume (Lr) of 4 L, were operated as semi-continuous stirred tank reactors (CSTR) at mesophilic conditions (37±1 ºC) during the whole experimental period. The temperature in each reactor was maintained using a thermoregulator (Ratek, TH7100) to circulate temperature controlled water through an external water jacket. The digesters were fed semi- continuously at 8h intervals using a peristaltic pump (Masterflex® L/S variable speed 100 rpm, pump head Masterflex® EasyLoad 3). Each digester was stirred using an overhead paddle stirrer set at 35 rpm (Heidolph, RZR 1). All digesters were equipped with a tipping bucket gas meter and gas collection system. The operation of the digesters is divided into two sections according to the research objectives.

Figure 10. Schematic representation of one digester set-up to evaluate the influence of inoculum source (4.2.1.1) and influence of retention time (4.2.1.2).

4.2.1.1 Determine influence of inoculum source The four digesters were inoculated each with a different inoculum source and operated identically for 295 days. The digesters were fed and drained simultaneously 3 times per day at a flow rate of 89 mL min-1. The four digesters were fed with synthetic media (section 4.1.2).

The start-up strategy was the same for all digesters. All the inocula were diluted with Milli-Q water to 10 g VS L-1 before inoculation. During the first 5 days, the digesters were fed once per day with -1 -1 an effective OLR of 0.25 g COD Lr d and a 40-day SRT. After day 5 the feeding strategy was -1 -1 changed to 3 times per day, increasing the OLR to 1 g COD Lr d . SRT was sequentially dropped from 36-day SRT (day 6 to 15) to 18-day SRT (day 16-34), and finally to 15-day SRT (day 35-295). The nominal operational conditions were solid retention time (SRT) of 15 days and an organic loading -1 -1 rate (OLR) of 1 g COD Lr d .

4.2.1.2 Determine the influence of retention time To isolate the SRT as a main pressure variable, the same four digesters running steady at -1 -1 15-day SRT and OLR of 1 g COD Lr d were subjected to a sequential SRT decrease: to 8-day SRT (day 296 – 333); 4-day SRT (day 334 – 373); and 2-day SRT (day 374 – 402) but maintaining a -1 -1 constant OLR of 1 g COD Lr d (days 296 – 402). Each SRT was operated for a minimum of 6 cycles and decreased when at least 3 cycles of steady data was obtained.

31 As in the previous 295 days of operation, the four digesters were fed with the same synthetic media (section 4.1.2). Feedstock concentration was adjusted at each experimental SRT to maintain a -1 -1 constant OLR (1 g COD Lr d ) through operation, therefore decoupling SRT from OLR. To avoid short-circuiting, effluent was first withdrawn from the digesters and feed was added after withdrawal was complete. Gas valves were installed in the gas line. The gas valves closed automatically during effluent withdrawal to avoid the entry of air into the system and opened after feeding was completed. Monitoring parameters included total solids, volatile solids, total and soluble chemical oxygen demand, alcohols and volatile fatty acid composition, pH, biogas production, and biogas composition (see section 4.3.1 for method description). Biogas production and biogas composition were measured daily. Digesters were analysed three times per week during the operation at 15-day SRT and 8-d SRT. Monitoring frequency was increased to four times per week at 4-day SRT and four to five times per week during the 2-day SRT operation.

4.1.2 Activity assays Batch activity assays were conducted on representative samples collected before the CSTR start- up (day 0), and at several times at 15-day SRT (day 90, 145, 185, and 295) and steady phase of each tested SRT (day 333, 373, 397). Activity assays were carried out at mesophilic temperature (37±1 ºC). Activity assays were performed in triplicate in 120 mL serum bottles, with a working volume of 80 mL. The serum bottles contained inoculum (as collected) and the amount of substrate required to achieve an initial inoculum to substrate ratio of 5 (VS-basis). The headspace of each bottle was −1 flushed with 99.9% N2 for one minute (4 L min ) and then sealed with rubber septa and aluminium caps. The bottles were placed in a temperature-controlled incubator. The bottles were mixed by swirling before each sampling event. The six model substrates to evaluate the relative microbial activities are shown in Table 1, representing the main AD trophic groups. Butyrate, propionate, acetate, and formate were added in the sodium salt form to avoid the initial pH drop caused by the acid forms.

To determine the hydrolytic rate, two sets of triplicates (six serum bottles) were run in parallel. One set was used to measure the soluble COD that accounts for the hydrolysed compounds that had not been methanised and the parallel set measures the methane production at the same sampling events. Since in AD systems COD is conserved, the amount of cellulose hydrolysed is proportional to the amount of COD solubilised and methanised. For the acetogenic activity assays, a sample of well-mixed liquor was extracted, centrifuged at 10.000x g for 5 min, and filtered (0.45 m PES Millipore® filter) at every sampling event to determine the substrate concentration. Methanogenic activity was tested directly using methane production, calculated from the pressure increase and

32 methane composition of the headspace at each sampling event. Methane production, in COD- equivalents, is reported at standard conditions (i.e. 0 ºC and 1 bar). Blank assays only containing inoculum were used to correct the background influence of the target substrate.

4.2.3 Fermentation batch assays Fermentation assays were carried out using primary sludge (PS) as a substrate (see section 4.1.2.3). 100 g of fresh PS were added to 300 mL glass bottles under semi-aerobic or anaerobic conditions. No inoculum was added to the fermentation assays. Semi-aerobic conditions were carried out by leaving the bottles open to the environment, while anaerobic conditions were ensured by flushing the headspace of bottles with N2 (99.9%) and sealing each bottle with a rubber septum and a screw cap. The first set of experiments was performed anaerobically at 20, 37, 55, and 70 ºC, and semi- aerobically at 20 ºC with four retention times at each temperature: 12, 24, 48 and 72 h. All experiments were performed without agitation. All tests were conducted in duplicate. The second set of experiments were performed under semi-aerobic conditions at 20 °C only with retention times of 24, 48, 72 and 96 h. The second set of semi-aerobic experiments was also performed in duplicate.

Destructive sampling was used, where serum bottles were discarded after each treatment time. For each test condition, the liquid fraction was separated by centrifuging the pre-fermented sludge at 2500x g for 5 min. Chemical analyses to determine the extent of solubilisation (soluble COD) and VFA production were done after filtering the liquid fraction through a 0.45 m PES Millipore® filter (see section 4.3.1.2). In the anaerobic fermentation tests, the headspace composition of each serum bottle was analysed just before processing the sample.

4.2.4 Biochemical methane potential test Biochemical methane potential (BMP) tests were carried out following the procedure defined by Angelidaki et al. (2009) at mesophilic temperature (37±1 ºC). BMP tests were performed in triplicate in 160 mL serum bottles, with a working volume of 100 mL, sealed with rubber septa and aluminium caps. The serum bottles contained inoculum and the amount of substrate required to achieve an initial inoculum to substrate ratio of 2 (VS-basis). Blank assays only containing inoculum were used to correct for the background methane potential of the inoculum. Next, the headspace of each bottle was -1 flushed with 99.9% N2 for one minute (4 L min ). Finally, the bottles were placed in an incubator set at 37 °C. Serum bottles were manually mixed by swirling before each sampling event. Accumulated volumetric methane production was calculated from the pressure increase and methane composition of the headspace at each sampling event. Methane yields are reported at standard conditions (i.e. 0 ºC and 1 bar).

Figure 11. Schematic representation of the second experimental set-up to determine the influence of operational conditions into a bio-refinery concept.

4.3 Methods 4.3.1 Analytical methods 4.3.1.2 Physico-chemical analyses Total solids (TS) and volatile solids (VS) were determined following the standard methods 2540G (APHA 2012). Total and soluble chemical oxygen demand (tCOD, sCOD), were measured using a Merck COD Spectroquant® test kit (range 0.5 – 10 g L-1 and range 25 – 1500 mg L-1 respectively) and a Move 100 colorimeter (Merck, Germany). pH was measured using an Orion™ Ross Ultra Electrode. Volatile fatty acids (VFA), also known as short-chain fatty acids, (acetate, propionate, butyrate, valerate, and caproate) and alcohols (i.e. methanol, ethanol, and butanol) were measured using an Agilent 7890A gas chromatograph equipped with an Agilent DB-FFAP capillary column and a flame ionisation detector. Biogas composition (H2, CH4 and CO2) was determined by gas chromatography using a Shimadzu GC-2014 equipped with an HAYESEPQ 80/100 packed column and a thermal conductivity detector; argon was used as a carrier gas.

4.3.1.2 Microbial analyses Genomic DNA was extracted from samples using the FastDNA™ SPIN kit for soil (MP Biomedicals, USA). Extractions were conducted according to the manufacturer's protocol. The V6 to V8 region of the 16S rRNA gene was amplified using the universal primers 926F (5’-

34 AAACTYAAAKGAATTGACGG-3’) and 1392R (5’-ACGGGCGGTGWGTRC-3’). As a quality control measure, DNA concentration and purity was determined by gel electrophoresis on 1% agarose gel, and spectrophotometrically using the NanoDrop ND-1000 (Thermo Fisher Scientific, USA). DNA was submitted to the Australian Centre for Ecogenomics (University of Queensland, Australia) for 16s rRNA Amplicon Sequencing. Sequencing was performed on the DNA using the Illumina MiSeq platform (Illumina, USA).

4.3.2 Data Analysis 4.3.2.1 Performance-related calculations AD yields The solubilisation yield, the acidification yield, and the methanogenic yield were calculated following Eq. 6, Eq. 7, and Eq. 8, respectively.

sCODfi- sCOD solubilisation yield = Eq. 6 tCODi

VFAfi- VFA acidification (VFA) yield = Eq. 7 tCODi

VCH methane yield = 4 Eq. 8 tCODi

-1 where sCODf and sCODi are the soluble COD (g COD L ) at the end and at the beginning of the experiment; and VFAf and VFAi are the total VFA concentration expressed in COD equivalents (g -1 COD L ) at the end and at the beginning of the experiment; VCH4 is the normalised (0 ºC, 1bar) methane production (LN); and tCODi is the initial amount of total COD added (g COD) Yields can be converted to VS-basis (initial concentration of VS) by multiplying by the COD:VS ratio of the substrate.

AD mass balance In AD processes there is a conservation of the COD, only being distributed. Therefore, at steady state, the digesters mass balance can be expressed in COD-basis as Eq. 9

q tCOD  COD  q  tCOD Eq. 9 influent influent CH4 effluent effluent

-1 where qinfluent is the influent flow rate (L d ), tCODinfluent is the substrate of total COD -1 concentration (g COD L ), CODCH4 is the daily methane production expressed in COD equivalents

35 -1 -1 (1g COD = 0.35 LNCH4) (g COD d ), qeffluent is the effluent flow rate (L d ), tCODeffluent is the substrate of total COD concentration (g COD L-1)

Overall methane yield The overall methane yield (B’) expresses the methane yield of the fermented sludge regarding the PS initial organic matter content (Eq. 10). B’ is used to normalise the PS methane yield by taking into account the organic matter losses occurring during waste processing (Astals et al. 2015).

BB'0  (1   ) Eq. 10

-1 where B’ is the overall methane yield (LCH4 kg VS), B0 is the methane yield of the waste after -1 the fermentation (LCH4 kg VS), and ρ is the organic matter losses expressed as per unit -1 (gVSfinal g VSinitial).

4.3.2.2 Activity assays and BMP modelling Activity assays Monod kinetics was used to describe the specific substrate consumption rate in the activity assays according to Soto et al. (Soto et al. 1993). Considering no biomass growth and an excess of substrate, Monod kinetics can be simplified to a zero order kinetics towards the substrate. Therefore, the -1 -1 maximum consumption rate constant, km (g COD g VSinoculum d ), is obtained as the slope of a linear regression fit (Analysis Toolpak in Microsoft Excel 2013) applied to the specific substrate -1 consumption on a COD-basis (g COD g VSinoculum)-(y-axis) and time (d)-(x-axis), for subsets of data over which the rate was approximately constant. The 95% confidence interval in slope was determined using a two-tailed t-test with n-2 degrees of freedom where n is the number of data points available for regression.

BMP and first-order kinetics The impact of fermentation and separation of soluble products on methane yield was assessed by mathematical analysis of the BMPs. Moreover, the hydrolytic activity assays were also modelled using first-order kinetics following Eq. 11 since hydrolysis was considered to be the rate-limiting step during the AD of PS and cellulose (Jensen et al. 2011).

r fi  k hyd, i  X i  C i Eq. 11

-1 where r is the methane production rate (L COD-CH4 d ), fi is the substrate biodegradability (-), -1 khyd,i is the first order hydrolysis rate constant of the substrate (d ), Xi is the substrate

36 -1 concentration (g VS L ), and Ci is the measured COD-to-VS ratio of the substrate -1 (COD:VS = 1.40 gCOD gVS ). To normalise and analyse model outputs, the biodegradability (fi) was estimated as per Eq. 12.

B0 fi  Eq. 12 B0,max

where B0 is the measured methane yield, and B0,max is the maximum theoretical methane yield at standard conditions (350 · COD:VS)

The model was implemented in Aquasim 2.1d. Parameter estimation and uncertainty analysis were simultaneously estimated, with a 95% confidence limit, as described in Batstone et al. (2009) and Jensen et al. (2011). Parameter uncertainty was estimated based on a two-tailed t-test on parameter standard error around the optimum, and non-linear confidence regions were tested to confirm the linear estimate was representative of true confidence. The objective function used was the sum of squared errors (χ2), where average data from triplicate experiments were used.

Digesters and first-order kinetics To describe the AD performance in the continuous digesters at each SRT, digesters were modelled using a simple first-order model at steady-state (accumulation = 0) and considering perfect mixing. Under these circumstances, the first-order equation can be derived to Eq. 13 (Garcia-Heras 2003) where AD process kinetics (k) is the target parameter.

k SRT BB Eq.13 0,max 1k SRT

-1 where B is the specific methane production yield (LNCH4 kgCODin ), B0,max is the maximum -1 theoretical methane yield at standard conditions (350 LNCH4 kgCODin ), k is the first-order AD process constant (d-1), and SRT is the solid retention time (d).

The model was implemented in Microsoft Excel 2013. Parameter estimation was carried out using a non-linear regression procedure based on Batstone et al., (2009). Briefly, the objective function used was the residual sum of squares between the experimental data and model. Appropriate F-values were used for the number of parameters and degrees of freedom in all cases.

37 4.3.2.3 Bioinformatics Raw DNA sequences were trimmed to 190 base pair length using Trimmomatic (Bolger et al. 2014). The forward and the assembled reads were independently examined to confirm the taxonomic composition; however, only the results from the forward reads were used in the subsequent analyses. Clustering was performed at 97% similarity using the QIIME v.1.9.0 pick_open_reference_otus.py script (Caporaso et al. 2010b) with the Greengenes database (v13_8). Alignment was performed with PyNAST (Caporaso et al. 2010a) and taxonomy was assigned using UCLUST (Edgar 2010), both with default QIIME parameters. The resulting OTU table was filtered only to retain those sequences that comprised 0.005% or more of the reads in at least one sample.

4.3.2.4 Statistical analyses Statistical analyses were performed in RStudio v3.3.1. Two-tailed t-test and ANOVA (5% significance threshold) were used to compare the digesters performance and activity rates between digesters and between SRTs. Microbial communities were explored using Vegan (Oksanen 2016) and Phyloseq packages (McMurdie and Holmes 2013). Differences in community composition and dynamics were explored by principal component analysis (PCA), where the OTU table was rarefied to the depth of the shallowest sample, using rarefy_even_depth() (rgn seed 712) from Phyloseq package. To accommodate the data to the statistical requirements of PCA analysis, the resulting rarefied OTU table was Hellinger transformed prior PCA. Correlations between microbial community composition and performance data were calculated using environmental parameter fitting envfit() in the package Vegan. Performance data was variance-stabilised (z-scoring) before correlation analyses to aid with the comparability of variables that have different magnitudes. -diversity was examined by Non-metrical Multidimensional Scaling (NMDS) based on Bray-Curtis dissimilarity using metaMDS() from Vegan package. Pearson and Spearman correlations between activity rates and taxa relative abundance were performed on Hellinger-transformed abundance and non-transformed calculated activity rates. Prior correlation test, variance due to primary factors (i.e. digester, time, or SRT) was subtracted.

Chapter 5

5 Deterministic mechanisms define anaerobic digestion microbiome and its functionality regardless of the initial microbial community

Partially in: Peces, M., Astals, S., Jensen, P.D., and Clarke, W.P. 2016. Do different inocula converge given the same operational conditions in long-term anaerobic digestion? In Proceedings: XII DAAL - Taller y Simposio Latino Americano en Digestión Anaerobia. Cusco, Peru. (Conference oral presentation) Peces, M., Astals, S., Jensen, P.D., and Clarke, W.P. 2017. Linking microbial populations and kinetic rates in anaerobic digestion. In proceedings: 15th World Congress on Anaerobic Digestion. Beijing, China. (Conference poster presentation – to be held October 2017)

39 5.1 Introduction Anaerobic digestion (AD) is a process used worldwide to treat and stabilise organic waste and currently stands as a key technology in the emerging green economy. AD is a biochemical microbially-mediated multi-step process where organic matter is hydrolysed, fermented, syntrophically oxidised to acetate and hydrogen, and converted to biogas (methane and carbon dioxide) in the absence of oxygen. Extensive process-based research and full-scale digesters operation have significantly improved the feasibility and reliability of AD. However, a better understanding of the associated microbial communities and their dynamics may reveal new avenues to further improve AD efficiency and broader application. Microbes in engineered AD systems are usually provided during the inoculation phase, where inoculum (e.g. digested sludge/manure) from a well-functioning anaerobic digester is transferred to a new digester. From an engineering perspective, the inoculum is seen as a source of active microbes to reach successful operation, while avoiding process failure during the start-up (Regueiro et al. 2012). However, given the key role of microbes in AD, it is fundamental to determine the impact of the inoculum microbial community on process performance, both during start-up and long-term operation.

The effect of microbial community composition on ecosystem functionality has primarily been assessed by determining the capabilities of the inoculum microbial community through short-term experiments (De Vrieze et al. 2015b, Langenheder et al. 2006, Perrotta et al. 2017). Although these tests provide an indicative response of the capabilities of the inoculum, they cannot predict if the characteristics will be maintained in the new environment or if the microbial community will be able to develop new capabilities. On the other hand, long-term continuous experiments have shown that a microbial community is governed, to a greater extent, by deterministic factors (Fernández et al. 1999, Lucas et al. 2015, Vanwonterghem et al. 2014a). However, it remains uncertain how different microbial communities will develop and function when subjected to the same operational conditions. Additionally, different microbial communities do not necessarily result in different process capabilities, as different microbes can be functionally redundant, allowing the ecosystem to maintain the same functionality with a dynamic population (Allison and Martiny 2008). In fact, unravelling the role of individual microbial groups, many of which can perform a variety of metabolic roles, remains a challenge (Martiny et al. 2015, Widder et al. 2016). However, even if individual taxa functioned differently, the combined performance of all taxa at a broad-scale level can be the same (Allison and Martiny 2008). It is evident that understanding the link between process performance, metabolic activity rates, and the microbial community has practical implications as different microbial communities could impact AD process efficiency, the end-product spectrum, and the response to environmental disturbances (Carballa et al. 2015, De Vrieze et al. 2015a). Therefore, in

40 engineered ecosystems such as AD where microbial communities are imported, it is critical to identify how the seeding microbial community and its characteristics will develop through the operation and whether or not that is reflected in process performance.

5.2 Aim and approach This study aims to evaluate how different anaerobic microbial communities adapt to certain operational conditions and how this affects long-term digestion performance. To achieve this aim, active microbial communities from four different full-scale anaerobic digesters (substrate type and reactor configuration) were each used to inoculate a continuous anaerobic digester. The four digesters were operated identically for 295 days on a feed of cellulose and casein. Digesters were monitored by (i) process performance, (ii) activity assays and (iii) microbial community composition.

In Figure 12 and Table 3 it is summarised the rationale, experimental set-up, and materials and methods to accomplish the aim of the present study. The full description of the materials and methods used in this study is described in Chapter 4.

Figure 12. Schematic representation of the rationale and experimental set-up used in Chapter 5

Table 3. List of materials and methods used in Chapter 5 previously described in Chapter 4 Materials Methods - Digestate slaughterhouse anaerobic lagoon (SL) - Continous digesters - Digested sewage sludge (SS) - Activity assays - Digestate piggery anaerobic lagoon (PL) - Analytical methods: - Anaerobic brewery granules (BG) o TS, VS, tCOD, sCOD, pH, VFA, Biogas - Synthetic feed media composition - Model substrates: o Microbial analyses Cellulose, butyrate, propionate, acetate, - Data analysis: formate o AD yields o Monod modelling o Bioinformatics o Statistical analyses

41 5.3 Results

5.3.1 Digester operation

Operation of the digesters was divided into 3 phases (Figure 13): (i) start-up (day 0 to 34), period where the SRT was decreased from 40 to 15 days; (ii) transition (day 35 to 80), period where the inoculum background was washed out (i.e. variable concentration of TS, VS and tCOD); and (iii) steady-state (day 81 to 295), period where all digesters run at stable performance. The operation of all digesters was disturbed due to a malfunction of the central pump controller on day 156 which caused an organic load shock (~4 times the nominal OLR) and exposed the contents of all digesters to air (~7 hours). All digester performance returned to pre-disturbance levels by day 165; data from day 157 to day 165 is not included in performance analysis. Biogas data from days 215 to 240 was not included in performance analysis due to a leak in the gas collection line. Table 3 and Figure A-II summarise the steady-state performance of the four digesters excluding the aforementioned periods. A

Figure 13. (A) Methane production yield of the 4 digesters (♦)SL, (♦)SS, (♦)PL, (♦)BG. Data within dotted red lines have been excluded from the data analysis due to operational complications. (B) VFA concentration. Operational stages correspond to the periods (■) Start-up, (■) Transition, and (■) Steady-State

42 The rapid establishment of methane production and low VFA concentrations indicated that the four inocula adapted to the new operational conditions, with the accumulation of propionate in PL (75 mg COD L-1) the only remarkable phenomenon. The steady-state conditions were reached from day 80 onwards, although activity assays profiles and microbial analysis shows that the community continued to evolve until days 120-145 (Figure 14, Figure 15). During steady-state conditions, digesters performance parameters were not significantly different. VFA levels averaged 19±1 mg COD L-1 (P > 0.250) being acetate the major VFA compound followed by traces of propionate, valerate, and butyrate. pH values were stable for the 4 digesters at 7.3±0.1 (P > 0.767). Methane yield -1 averaged 279±22 LNCH4 kgCODin (P > 0.783), which corresponds to a feedstock COD removal of 80%. This value is in agreement with the COD removals values (80±1%, P > 0.178) calculated by measuring the influent and effluent COD (Table 4). The latter proofs that the measured data is both consistent and representative of process performance (Figure A-III for COD balance).

Table 4. Effluent characteristics and process efficiency for each digester at steady state conditions (average ± standard deviation)* SL SS PL BG TS (g L-1) 6.0 ± 0.3 6.1 ± 0.3 6.1 ± 0.2 6.1 ± 0.4 VS (g L-1) 2.8 ± 0.2 2.9 ± 0.3 2.9 ± 0.2 2.9 ± 0.3 tCOD (g COD L-1) 3.0 ± 0.2 2.9 ± 0.2 3.0 ± 0.3 3.0 ± 0.2 sCOD (mg COD L-1) 218 ± 77 243 ± 78 219 ± 108 312 ± 81 VFA (mg COD L-1) 18 ± 5 19 ± 6 19 ± 4 21 ± 6 pH 7.3 ± 0.1 7.3 ± 0.1 7.3 ± 0.1 7.3 ± 0.1 Methane yield 277 ± 21 280 ± 23 278 ± 26 281 ± 22 -1 (L CH4 kgCODin )

CH4 content (%) 53 ± 2 53 ± 2 53 ± 1 53 ± 1

CODremoval (%) 79 ± 3 81 ± 3 80 ± 3 80 ± 2

* number of samples used to calculate average: TS, VS (n = 32); pH, sCOD, tCOD, VFA (n = 36); methane yield (n = 179)

5.3.2 Activity assays Activity assays along the experimental period were used to evaluate the capacity of the main AD metabolic steps (i.e. hydrolysis, acetogenesis, and methanogenesis). The consumption rate of cellulose, butyrate, propionate, acetate and formate reflects the ability of the microbial community to utilise the specific substrate.

The substrate consumption rates in response to all substrates over time are shown in Figure 2. The initial activity assays (day 0) were carried out before digester start-up and reflect the behaviour of the

43 inocula in their native habitats. The initial activity profile of each inoculum was different. SL and BG showed higher acetogenesis (both butyrate and propionate) and acetoclastic methanogenesis rates compared to SS and PL. The hydrolytic and hydrogenotrophic methanogenesis rates were more similar among inocula. Additionally, each inoculum had a singular balance between metabolic steps (Figure A-IV).

B C

D E

44 The activity rates of the tested substrates converged among digesters after 145 days of operation, -1 -1 with averages of: (i) 0.08±0.01 gCOD g VSinoculum d for cellulose hydrolysis; -1 -1 -1 -1 (ii) 0.19±0.02 gCOD g VSinoculum d for butyrate degradation; (iii) 0.18±0.02 gCOD g VSinoculum d -1 -1 for propionate degradation; (iv) 0.18±0.03 gCOD g VSinoculum d for acetoclastic methanogenesis; -1 -1 and (v) 0.12±0.02 gCOD g VSinoculum d for hydrogenotrophic methanogenesis.

5.3.3 Microbial communities The development of the seeding microbial communities was evaluated by PCA ordination based on Hellinger transformed rarefied abundances (Figure 15) and NMDS ordinations based on Bray- Curtis dissimilarity (Figure A-V). Both analyses showed that the trajectories traced by the four inocula migrated to a similar region in the ordination space after 120-145 days of operation (about 65 days after physicochemical convergence), with time as the only significant factor (P=0.001). Different groups of taxa increased in abundance with operational time while others were progressively washed out from the digesters as shown in Figure 16 and Figure A-VIII. Among the initially dominant groups, Methanobacterium, Methanolinea, various representatives from phylum Chloroflexi (Anaerolinaceae, and Anaerolinaceae candidate genus T78), Firmicutes (Bacilli and Clostridiales), candidate phylum OD1, and Synergistetes (genus E6) decreased in abundance. The groups that flourished in the digesters and formed a core community were Methanomicrobiales (Methanoregulaceae and Methanospirillium), Ruminoccocaceae, Spirochaetes (Treponema and Sediment-4), Kosmotoga and Dethiosulfovibrionaceae (candidate genus HA73). OTUs belonging Methanosaeta and Bacteroidales were found in high abundance (average relative abundance 16.7±7.3% and 10.0±5.3%, respectively) at all times in all digesters. In addition to microbial community composition, the estimated richness also showed that the microbial structure converged for the four digesters to an average of 358±15 16S OTUs. Similarly, Simpson diversity (D) and Shannon-Wiener (H’) entropy showed the synchronisation of all digesters after ~185 days with averages of 0.95±0.01 (D) and 4.05±0.19 (H’) (see Figure A-X).

The four starting inocula shared an average of 31% of the OTUs representing around 43% of the total relative abundance, with bacteria covering less than one-fifth of the initial shared community (Figure A-VI). During the microbial steady-state (day 120 – 295) 52% of all identified OTUs were common to all digesters, and this core community accounted for 72% of the total microbial community relative abundance (Figure A-VI), where about three-fourths of the shared community belonged to bacteria. Additionally, the procrustes analyses (Figure A-VII, Table A-I) showed that the microbial community development of the four digesters followed a coordinated pattern of change despite the initial microbial composition.

Figure 15. PCA of the microbial community profiles at the OTU level (Hellinger transformed) for the four digesters over 10 sampling events (day 0 to 295). Each sample is represented by a single circle, coloured according to digester designations in previous plots. Circle size increases with time. OTUs are presented as black crosses, and the populations contributing most to the variability between microbial communities are identified. Correlation with performance parameters is indicated by the arrow.

Figure 16. Microbial composition at 97% similarity. The heatmap shows the populations (>2.5% relative abundance in at least one sample) in the four digesters at the 10 sampling points. The lowest possible taxonomic assignment is shown in the right column and phylum level in the left column. Darker intensity indicates higher relative abundance; grey cells indicate that these taxa were not detected in the samples at the level of resolution (>0.01% of total counts at least in one sample).

5.3.4 Linkage of microbial composition to functionality The combination of 16S rRNA with metabolic activity assays outputs was used to correlate microbial populations to process functionality. Some relationships between taxa (at 97% similarity, using the lowest possible taxonomic assignment) and activity rates were identified. Figure 17 and Figure A-IX show Pearson and Spearman correlations, where only positive correlations have been considered. The presence of Flavobacteriaceae, representatives of Chloroflexi, Firmicutes, candidates phyla OD1 and OP8, and Armatimonadetes were positively correlated with an increment

47 in cellulolytic rates; however, these groups belonged to taxa with relative abundances of less than 2.5% in all samples (see Figure A-VIII). Butyrate consumption rate was positively associated with Syntrophomonas, Synthophobacter, and Bacteroidete (BA008), while propionate consumption rate was positively correlated with Clostridium, Veillonellaceae, Succiniclasticum, Phycisphaeraceae and Syntrophobacter.

Figure 17. Correlation maps display the taxa that showed a significant positive correlation with at least one metabolic rate. Left, Pearson correlation taxa- metabolism (linear). Right, Spearman correlation taxa-metabolism (monotonic). Blue represents positive correlations, red represents negative correlations with taxa abundance. Colour intensity represents the strength of the correlation (darker, stronger). Numbered cells indicate those where the correlation coefficient is significant (Bold, P <0.01, Italic P < 0.05).

48 5.4 Discussion

5.4.1 Microbial composition linked to process function and activity rates 5.4.1.1 Process and microbial community performance. AD efficiency is governed by the limiting step (the slowest). As shown by the activity assays results (Figure 14), hydrolysis was the limiting step regardless of the inoculum or period of operation in the continuous digesters, being ~2 times slower than the second slowest step (see Figure A-IV). This is consistent with the literature where hydrolysis is generally accepted as the limiting step when treating particulate organic matter (Vavilin et al. 2008) and explains the low VFA concentrations (<20 mgCOD L-1) detected in all digesters. Methane production of the 4 digesters converged after 60 days (~3.5-SRT, Figure 13A). However, neither the activity rates nor the microbial communities had converged by this stage (Figure 14, Figure 15). This can be explained by the impact of the limiting step (i.e. hydrolysis) on methane production. Indeed, hydrolysis was not only the slowest step but also the activity rate that fluctuated the least over time (Figure 14). The stable hydrolysis rate combined with a relatively long SRT (15 days) caused the convergence of the methane production regardless of the microbial population. The fact that different microbial communities exhibited similar hydrolysis rates indicates functional redundancy of the hydrolytic population (Carballa et al. 2015). Using functional assignments based on literature, the hydrolytic population of SL and SS was initially dominated by Chloroflexi, WWE1 and Clostridium, PL inoculum by Clostridiales and Bacilli, while BG inoculum had a diverse hydrolytic population but with lower relative abundance than the other inocula (Figure 16). The potential hydrolytic population of the 4 digesters converged after 145 days and was dominated by Rumminoccocacaceae, Bacteroidales, Thermotogales, and Spirochaetes.

Although process efficiency was controlled by the hydrolytic step, AD is a highly interdependent process and, therefore, the other steps (i.e. acetogenesis and methanogenesis) should also be considered. Acetogenesis was in most cases the fastest step as well as the step with the most variable activity rate (Figure 14). As a general trend, the relative abundance of Syntrophaceae, Syntrophobacteraceae, Syntrophorhabdaceae, and Dethiosulfovibrionaceae stabilised to similar levels in the 4 digesters after >120 days. Acetoclastic methanogenesis was faster than hydrogenotrophic methanogenesis in the four digesters through most of the experiment suggesting that acetoclastic methanogenesis is the preferred methanogenic pathway at these operational conditions. Concomitantly, Methanosaeta, an acetoclastic methanogen, was the dominant methanogen in all digesters for the full duration of the experiment. The other methanogens present in the digesters were identified as hydrogenotrophic methanogens, which displayed a higher variability than the acetoclastic population. Methanobacteriales (present in all native inocula but with different abundance) vanished in favour of Methanomicrobiales, although this shift in composition and

49 abundance did not lead to changes in the formate consumption rate. This suggests a high level of functional redundancy within hydrogenotrophic methanogens. The presence of Clostridium and Synergistetes indicates that syntrophic acetate oxidation (SAO) pathway could have occurred in the digesters as some representatives of these taxa have been reported as SAO bacteria (Hao et al. 2015, Ito et al. 2011, Mosbaek et al. 2016)

5.4.1.2 Individual microbial populations associated with activity rates Correlation results related higher hydrolysis rates with microbes such as Flavobacteriaceae, Chloroflexi candidate class SHD-231, Firmicutes candidate class SHA-98 and other bacterial phyla which are still not fully understood, such as candidate phylum OP8 or OD1. The aforementioned bacteria tend to be very diverse and associated with several capabilities as the use of polysaccharides, the conversion of high molecular weight compounds to low molecular weight ones, and show glycolysis metabolism (Buchan et al. 2014, Farag et al. 2014, Nelson and Stegen 2015, Sun et al. 2015, Sun et al. 2016). For instance, the candidate class SHA-37, from the recently defined phylum Armatimonadetes showed a strong positive correlation with the hydrolysis rate. Armatimonadetes are defined as aerobic soil-based oligotrophs (Lee et al. 2014); however, Armatimonadetes have been identified, in low abundance, in several AD processes (Li et al. 2015, St-Pierre and Wright 2014, Tuan et al. 2014). These previous observations indicate that within Armatimonadetes some strains are not strict aerobes, although their role in AD is to date unidentified. The present results suggest a conceivable involvement of the candidate class SHA-37 in the hydrolysis-fermentation step. The correlation analysis does not allow discriminating if the forenamed bacteria are directly involved in the cellulose break down or utilising the monomers released from the hydrolysis (Berlemont et al. 2014); however, the presence of these bacteria benefited the hydrolysis.

Acetogenic consumption rates showed positive correlations with well-known acetogenic bacteria such as Syntrophomonas (butyrate consumption) and Syntrophobacter (butyrate and propionate consumption). Veillonellaceae, which can ferment sugars to propionate via acrylate and succinate pathways (Louis et al. 2014), showed a positive relationship with the propionate consumption rate. This suggests that propionate-consuming organisms are triggered by propionate supply and demonstrates the important interactions between metabolic steps in AD. Similarly, the candidate family BA008, associated with Bacteroidetes, could have been contributing to butyrate production supporting the higher rates of butyrate consumers. There is currently little information available on T BA008. However the only strain isolated to date, TBC1 , has an anaerobic fermentative metabolism (Sun et al. 2016). Methanogenic rates could not be linked to shifts in the archaeal community since one of the variables was relatively constant. On the one hand, Methanosaeta was ubiquitous, but acetate

50 consumption rates were variable. On the other hand, the relative abundance of hydrogenotrophic methanogens was dynamic, but similar formate consumption rates were obtained. However, the positive correlation of taxa able to produce acetate as an end product, such as Syntrophobacter, Geobacter and some Bacteroidales (Kuever 2014, Rotaru et al. 2014), further supports the concept that metabolic processes are interdependent. Remarkably, Fibrobacter showed a positive correlation to the formate consumption rate. The complete genome of Fibrobacter Succinogenes revealed a specialised cellulose hydrolyser that utilises the soluble hydrolysis products and produces formate as one of the major fermentative products (Suen et al. 2011). Pirellulaceae has a subset of genes encoding the C1 carbon metabolism; however, the ability to use or produce C1 compounds has not been yet observed (Fuerst and Sagulenko 2011). While this test does not provide conclusive evidence, the association is prominent.

This correlation study showed some linkages between microbial communities and process functioning. Nonetheless, these must be interpreted and extrapolated with caution. Spurious correlations due to biases in the determination of microbial relative abundance, normalisation of activity rates, or merely by chance cannot be discarded. Further experiments must be carried out to complement and confirm the relationships presented in this study.

5.4.2 Development of microbial communities towards a core-community The four starting inocula were collected from healthy AD full-scale plants. These AD processes differed in substrate type, organic and hydraulic load, process configuration and temperature. Therefore, the chosen laboratory digesters conditions represented an operational shock to the native communities. The most relevant differences between the native and the operational laboratory conditions (substrate type, biomass retention time, and mixing regime) were the main deterministic contributors. By identically operating the four digesters, the major source of stochasticity was attributed to the initial microbial communities.

The PCA ordination (Figure 15) and the analysis of the procrustes rotations (Figure A-VII, Table A-VII) showed that the four starting microbial communities followed similar successional trajectories from day 0 until the end of the experimentation. The convergence of the different microbial communities towards a core-community proves that the deterministic factors (process operational conditions) were a stronger driver than the initial microbial community composition. The development of a core-community representing a 72% of the relative abundance of the final microbial communities (Figure S-VI) is in agreement with recent findings that concluded that a bioreactor microbial community is mainly regulated by deterministic processes rather than stochastic (Valentin- Vargas 2012 #121) (Lin et al. 2017, Luo et al. 2015, Vanwonterghem et al. 2014a). However, even

51 after 295 days, each digester retained a significant number of unique OTUs, mostly bacteria, explaining up to a 5% of the relative abundance indicating a contribution of stochastic deviations. Nonetheless, these differences (attributable to the original seed inoculum) did not contribute to a difference in functionality as shown by digesters performance and the activity rates profile.

Finally, the patterns observed among the development of the four microbial communities strengthen the concept of synchrony. Synchronised patterns have been reported in natural habitats such aquatic ecosystems (Kent et al. 2007) and animal gut microbiota (Bergmann et al. 2015), and engineered bioreactors such as activated sludge bioreactors (Griffin and Wells 2017), denitrifying bioreactors (Porter et al. 2015), and replicated AD digesters seeded with the same inoculum (Vanwonterghem et al. 2014a). The novelty of this work is that synchronisation and convergence were observed even when the digesters were started with a different microbial community.

5.5 Conclusions The operation of four identical anaerobic digesters each started with a different inoculum has demonstrated that long-term process performance is independent of the starting microbial community composition, with the limiting step (i.e. hydrolysis) dominating process efficiency. After inoculation, digesters performance converged and stabilised in 80 days, while activity rates and microbial communities converged and stabilised after 145 days of operation. The synchronisation of the microbial communities and the development of a core-community strengthens the hypothesis that microbial communities and their functionality are mainly driven by deterministic factors as a response to specific operational conditions. Hence, a desired microbial population could be achieved and predicted by tunning process conditions. Finally, the positive correlation of activity rates with microbial abundance has identified several bacterial taxa with a conceivable involvement in the different anaerobic digestion steps, expanding the understanding of microbial communities and functionality.

Chapter 6

6 Transition of microbial communities and degradation pathways in anaerobic digestion at decreasing retention time

53 6.1 Introduction Anaerobic digestion (AD) stands as a key biotechnology in the emerging green economy due to its capacity to convert organic waste into renewable biogas energy {Batstone, 2014 #249}. In more advanced applications, anaerobic processes can be engineered to produce volatile fatty acids (VFA), alcohols, and hydrogen which are precursors to a range of bio-based products such as biofuels, bioplastics and biotextiles {Agler, 2012 #244; Puyol, 2017 #200}. However, one of the key challenges of mixed culture anaerobic systems is to drive the process towards a specific product yield and profile {Hoezle, 2014 #201; Oleskowicz-Popiel, 2018 #202}. The microbiology of AD involves a large phylogenetic diversity of bacteria and archaea that have different functionalities, activity (kinetic) rates, growth rates, and nutritional requirements (Carballa et al. 2015, De Vrieze et al. 2017). In general, bacteria hydrolyse and ferment organic matter to VFA, alcohols and hydrogen, while archaea, mainly methanogens, convert acetate and hydrogen to methane. Consequently, the first step to increase the fermentation product yield is to inhibit or washout the methanogens from the process. Moreover, due to the large diversity of bacteria and fermentation metabolic pathways a better understanding of the link between operational conditions, microbial communities and product profile is required.

Directing the AD process towards a specific product can be achieved by manipulating the operational conditions such as pH, temperature, organic loading rate (OLR), and solid retention time (SRT). These manipulations influence both process performance and microbial communities to different extents (Heyer et al. 2016, Hoelzle et al. 2014). SRT is one of the easier process variables to manipulate. In continuous stirred tank reactors (CSTR), SRT represents the average time that particles (microbial populations or solid substrates) remain in the digester. SRT can affect the extent of organic matter removal, either through the time available for degradation or through the microbial communities able to grow. Moreover, decreasing the SRT usually implies a higher OLR, which means that microbes receive more substrate per unit of time. The combination of a shorter SRT and the higher OLR in AD has been widely studied, which results in the accumulation of VFA and hydrogen, a decrease in the pH, and a decrease of the methane production (Dareioti et al. 2014, Nges and Liu 2010). The impact of OLR at a fixed SRT has been largely studied in AD, as a tool to optimise the methane production (Mata-Alvarez et al. 2014).In contrast, there is a lack of understanding of the individual effect of SRT as a tool to manipulate AD product profile. In this scenario, by varying SRT, it can be evaluated the impact of the selection pressure on microbial community composition and the response in process performance, stability, and product profile.

54 6.2 Aim and approach The aim of this study is to determine the effect of SRT, as an individual selective pressure parameter on AD. Four digesters operating at steady conditions at an SRT of 15 days for more than 200 days (Chapter 5) were subjected to a sequential decrease in SRT from 15 to 8 to 4 and 2 days while maintaining a constant OLR. Digesters evolution was closely monitored by (i) process performance (methane and VFA yields), (ii) activity assays (hydrolytic, acetogenic, and methanogenic rates) and (iii) microbial community composition (identity and relative abundance).

In Figure 18 and Table 5 it is summarised the rationale, experimental set-up, and materials and methods to accomplish the aim of the present study. The full description of the materials and methods used in this study is described in Chapter 4.

Figure 18. Schematic representation of the rationale, experimental set-up used in Chapter 6

Table 5. List of materials and methods used in Chapter 6 previously described in Chapter 4 Materials Methods - Continous digesters operated at 15-d SRT - Continous digesters (Chapter 5) - Activity assays - Synthetic feed media (cellulose:casein) - Analytical methods: - Model substrates: o TS, VS, tCOD, sCOD, pH, VFA, Biogas o Cellulose, butyrate, ethanol, propionate, composition acetate, and formate o Microbial analyses - Data analysis: o AD yields o Monod modelling o First-order modelling o Bioinformatics o Statistical analyses

55 6.3 Results

6.3.1 Digester operation Digesters operation was divided into four stages, one for each different SRT (Table B-I). Figure 19 shows the methane yield and VFA concentration of the four digesters over time. After decreasing the SRT, the digesters achieved a new steady state within 3-4 SRT equivalent cycles. This phenomenon was repeated at all SRT, showing that stable operation (defined as variations in methane yield and VFA yield <10%), could be achieved at SRTs as low as 2 days. An operational disturbance occurred at the beginning of the 4-d SRT stage (day 346) where digesters SS, PL and BG were not fed, but drained, for 1.5 days. This situation caused the exposure of the digesters content to air. However, as can be seen in Figure 19A, the three affected digesters recovered rapidly the methane yield following the similar trend than SL. The operational disturbance did not occur to SL. All data from this exposure period have been excluded from the analysis.

All digesters showed a progressive decrease in methane yield as the SRT was shortened, concomitant to a decrease in COD removal (Figure B-III). Methane production yields were not significantly different between digesters at 15-d SRT and 8-d SRT (P15-d SRT > 0.163, P8-d SRT > 0.137). However, SS diverged from SL, PL and BG producing significantly less methane at 4-d SRT and -9 -10 2-d SRT (P4-d SRT < 6.6·10 , P2-d SRT < 6.3·10 ). This was consistent with less COD removal in SS (Figure B-III), indicating that the lower methane production yield was characteristic of this digester.

-1 In all digesters the normalised VFA profile (g CODeq kg CODin ) was dominated by acetate and propionate followed by minor concentrations of i-butyrate, i-valerate, and n-butyrate (Figure B-I and Figure B-II). VFA accumulation tended to increase as the SRT was shortened with the transition from 8-d SRT to 4-d SRT the SRT having the highest impact on VFA accumulation and profile (Figure 1B). Despite the increase in VFA production at the shorter SRTs (4-d SRT and 2-d SRT) VFA concentration contributed to less than 3.5% of the influent COD. Interestingly, no statistically significant differences in VFA profile (yield, concentration and distribution) were observed between digesters at 15-d SRT and 8-d SRT, while differences between digesters were detected at 4-d SRT and 2-d SRT (Figure B-I, Figure B-II and Figure B-III).

Finally, the COD balance (Figure B-IV) was closed at 15-d, 8-d and 4-d SRT (except at 4-d SRT for BG), while at 2-d SRT around a 20% mistmach was calculated, where COD removed was higher than the methane yield obtained, suggesting that the four digesters were not behaving as model CSTRs. However, methane production in the four digesters was well represented by a first-order kinetic model (Figure B-IXA) indicating that the discrepacy was more related with the COD

56 measurement. In fact, the COD balance mistmach coincided with the formation of a biofilm on the paddles and walls of the digesters, causing a biomass (COD) accumulation.

Figure 19. (A) Methane production yield of the 4 digesters (♦) SL, (♦) SS, (♦) PL, (♦) BG expressed in COD-equivalents (1g COD = 0.35 LNCH4). (B) Normalised VFA concentration (VFA yield). Data within dotted red lines have been excluded from the data analysis due to operational complications.

57 6.3.2 Activity assays The activity rate of the six model substrates representing three main metabolic steps under study (i.e. hydrolysis, acetogenesis and methanogenesis) are shown in Figure 20. Cellulose hydrolysis and hydrogenotrophic methanogenesis (HM) were independent of the SRT (Phyd>0.5657, PHM>0.1959), whereas the acetogenic activity (butyrate, propionate, and ethanol) and acetoclastic methanogenesis -6 -5 (AM) were dependent on the SRT (PBu<0.0097, PEt<2.3·10 , PPro<0.0438, PAM<8.9·10 ). Propionate and butyrate activities showed a gradual decrease as the SRT was shortened; in contrast, ethanol activity increased as SRT was shortened. This result proves a shift in fermentation pathways depending on SRT. AM activity was clustered into two groups depending on the SRT. The average -1 -1 AM activity rate for the digesters comprised ranges: (i) 0.21 ± 0.02 gCOD gVSinoc d (15-d and 8- -1 -1 d SRT), and (ii) 0.08 ± 0.04 gCOD gVSinoc d (4-d and 2-d SRT). However, at 2-d SRT no AM activity was detected in the SS effluent, although a steady rate of methane was generated in the parent reactor, suggesting acetoclastic methanogenesis in SS was exclusively occurring in the biofilm.

The ratio between activity rates was also influenced by the operational SRT (Figure B-V). Cellulose hydrolysis was the slowest step (i.e. limiting step) at 15-d and 8-d SRT. However, AM (except in digester SL) and propionate activity were slower at 4-d and 2-d SRT. This change in the metabolic balance explains the presence of acetate and propionate in the effluent of the parent digesters, suggesting that AM and propionate degradation became rate limiting steps. The ratio between HM and AM (km,fo/km,ac; Figure SII-i) also shifted from an average of 0.56 ± 0.08 at 15-d and 8-d SRT to an average of more than 1.45 at 4-d and 2-d SRT, showing the influence of the SRT on the methane generation routes.

58 A B

B C

D E

Figure 20. Evolution of activity rates (km) over time, the solid coloured line represents the substrate consumption rate for each digester (–) SL, (–) SS, (–) PL, (–) BG. Shadowed coloured box represents the 95% confidence interval of the activity rate value.

6.3.3 Microbial communities 6.3.3.1 Influence of SRT The main factor influencing microbial communities was the SRT, as represented by the longer length of the arrow on the PCA plot (Figure 21). However, the SRT did not apply enough pressure to washout entire phyla, but caused changes at lower taxonomic levels (e.g. family or genus) as shown in Figure 22 (and Figure B-IX).

Figure 21. PCA of the microbial community profiles at the OTU level (Hellinger transformed) for the four digesters over 16 sampling events (4 at each SRT). Each sample is represented by a single circle, coloured according to digester designations in previous plots. Circle size decreases for each SRT. Squares represent the communities of the biofilms in each digester at the end of the experiment. OTUs are presented as black crosses, and the populations contributing most to the variability between microbial communities are identified. Correlation with performance parameters are indicated by arrows; blue arrows represent those parameters correlated with a significance lower than 0.01, and orange arrows parameters with a significance lower than 0.05

Bacteroidetes, Chloroflexi, Firmicutes, Spirochaetes, and Thermotogales were the most abundant bacterial phyla potentially associated with the hydrolytic-fermentative step. Within Bacteroidetes, genus Sphingobacterium and candidate genus Bliivi28 increased in relative abundance at lower SRTs, while an unclassified Bacteroidales decreased in relative abundance. A similar change in Thermotogales was observed, where the dominance of the Kosmotoga genus at long SRTs shifted to the candidate AUTHM297 at shorter SRTs. The family Anaerolineaceae (Chloroflexi), showed 2 significant OTUs for SL, SS and BG at 4-d and 2-d SRT. In all digesters, the genus Treponema (Spirochaetes) dominated at the shorter SRTs (4-d and 2-d). Firmicutes were less affected by the change in SRT across digesters except for the family Ruminococcaceae. Ruminococcaceae, was dominant in all digesters at long SRTs, but only remained dominant in SL digester at short SRTs, with 6 OTUs belonging to this family dominating the microbial community at 4-d and 2-d SRT (30±6% of the relative abundance).

Figure 22. Microbial composition at 97% similarity. Heatmap shows the suspended biomass populations (>2 % relative abundance at least in one sample) in the four digesters at the 16 sampling points and the biofilm at the end of the experiment. The lowest possible taxonomic assignment is shown in the right column and phylum level in the left column. Darker intensity indicates higher relative abundance; grey cells indicate taxa not detected in the samples at the level of resolution (>0.005% of total counts at least in one sample)

Bacteria potentially associated with acetogenic activity such as Syntrophus, Syntrophobacteraceae, Syntrophorhabdaceae, and Geobacter (phylum Proteobacteria) were strongly correlated to SRT (Pearson  > 0.37, P < 0.006), decreasing in relative abundance as the SRT was decreased. The two genera of the family Dethiosulfovibrionaceae (phylum Synergistetes) switched dominance with SRT. Candidate genus HA73 dominated at long SRTs (15-d and 8-d) and candidate genus PD-UASB-13 was found at the lower SRT (4-d and 2-d).

Methanogens were dominated by the genus Methanosaeta in all the digesters, and the relative abundance of this genus was not influenced by the SRT (Pearson  = 0.14, P > 0.272) (Figure 22). However, at the OTU level it can be observed that Methanosaeta was comprised of 5 major OTUs and that only one those OTU significant at 2-d SRT (Figure B-VIIIB). For the hydrogenotrophic methanogens, Methanobacterium tended to increase in relative abundance with lower SRTs, while Methanospirillium, Candidatus Methanoregula, and Methanoculleus were more abundant at 15-d and 8-d SRT (Figure B-VIIIB).

61 6.3.3.2 Influence of the operational disturbance The operational disturbance that occurred SS, PL and BG at the beginning of 4-d SRT influenced the microbial composition of those digesters. As can be seen in Figure 21, the microbial communities of SS, PL and BG diverged from the undisturbed digester (SL) at 4-d SRT and 2-d SRT migrating differently in the PC2-axis. The difference in community structure was mainly driven by the higher relative abundance of Rumminococaceae (3 different OTUs) and Rumminoccocus in SL. The divergence point coincides with the operational disturbance that affected SS, PL, and BG with the disturbance appearing to favour the establishment of different microbial communities. However, as shown in Figure 19 and Figure 20, the operational disturbance did not affect the overall process performance, nor did the disturbance cause a major difference in activity rates between digesters, indicating that functional redundancy occurs in anaerobic microbial ecosystems.

6.3.3.3 Biofilm formation Biofilm growth was observed on the walls and mixing paddle of all digesters with the decrease in SRT, probably as a stress response due to the short SRT. Biofilm formation was not synchronised in all digesters but occurred at different stages and to a different extent (Figure B-X). The formation of biofilm was visually documented around day 360 (4-d SRT) for BG, at the early stages of 2-d SRT for SL and PL, and by the end of the experimentation time for SS. It is worth mentioning that BG was originally inoculated with granular sludge, whereas SL and PL were collected from anaerobic lagoons and SS from a complete mixed digester. Despite the long operation since inoculation, and the high degree of similarity achieved by the microbial communities, based on 16S rRNA sequencing, in the four digesters over time, BG could have retained some organisms that were more readily prediposed to form a biolfilm when SRT was decreased.

The composition of the microbial communities in the digesters biofilms showed a higher similarity with the suspended microbial communities during the period where the biofilm formation commenced (Figure 21). However, the comparisons between digesters showed that biofilm microbial communities shared low similarity across the digesters (average 28.5±5.5% as opposed to the 47.7±3.0% calculated for the suspended biomass at 2-d SRT; inverted Bray-Curtis index at OTU level). While the four biofilm communities showed a diverse microbial community (Figure 22), the most common taxa (>1% relative abundance) shared in all the biofilms included Methanosaeta (>10% relative abundance), Bacteroidales (>3.5% relative abundance), Treponema (>2.2% relative abundance), and Anaerolineaceae. (>1% relative abundance). It seems likely that these taxa were involved in the biofilm formation and its functionality.

62 Finally, the periods corresponding to biofilm formation coincide with the unexpectedly low VFA yields measured in BG at 4-d SRT compared to SL, PL and SS (Figure 19B), and around the period when the COD balance was not closed for all the digesters. This suggests that biofilm communities (e.g. methanogens) had a plausible role in maintaining the relative low VFA concentrations and high methane yields at low SRTs.

6.4 Discussion

6.4.1 Influence of SRT on AD process performance The decrease in methane yield as the SRT lowered was well represented by a first-order kinetic -1 model, with optimum first-order kinetic constant (kh) values ranging from 0.25 to 0.40 d (see Figure

B-XIB). The kh values obtained from the digesters were similar to the kh values obtained in the hydrolytic activity assays, indicating that the rates obtained from the activity assays were representative of the continuous AD process. A close evaluation of the activity rates further indicated that the Monod hydrolysis activity (km) rates were constant regardless of the operational SRT. However, there was a decline in the propionate, butyrate and AM activity rates as the SRT decreased (Figure 20). At shorter SRT the propionate, butyrate and AM activity rate reduced to values similar to the hydrolysis activity rate. Consequently, the AD process performance transitioned from being exclusively limited by the hydrolysis step at 15-d SRT and 8-d SRT to be co-limited by hydrolysis and acetogenesis or acetoclastic methanogenesis at 4-d SRT and 2-d SRT (Figure B-V). Despite the transition of the activity rates, even at SRTs as low as 4 and 2 days, methanogenesis was active enough to maintain low concentrations of VFA (not exceeding 3.5% of the influent COD) and avoid digesters acidification (Figure 19B). However, it is important to highlight that in the present study the SRT was uncoupled from the OLR, thus isolating the individual effect of SRT on process -1 -1 performance. Maintaining the OLR at 1 gCOD Lr d (i.e. vary substrate concentration) prevented the general trend observed in the literature, where lowering the SRT causes an accumulation of VFA, a decrease in pH, and eventual cesation of methanogenic activity (Ho et al. 2013, Nges and Liu 2010, Vanwonterghem et al. 2015) probably because of the unbalanced growth between fermenters and methanogens.

The formation of biofilm on the digester walls and paddles reduced the washout of key microbes and facilitated steady methane production at 2-d SRT. This phenomenon is strongly supported by the batch activity tests where no acetoclastic methanogenic activity was detected for the SS sample during batch testing, however the continuous digester was steadily producing methane (the acetoclastic methanogenesis activity produced no methane, and in the hydrolytic assay all the solubilised cellulose accumulated as soluble COD). This result suggests that the active acetoclastic methanogens in the SS

63 during 2-d SRT were attached to the biofilm and not in the suspended medium. Further supporting this hypothesis is that the relative abundance of Methanosaeta in the biofilm composed 44% of the microbial community, as opposed to an average relative abundance of 12% in suspension.

6.4.2 Activity rates inferred a switch in degradation pathways Given the dominance of the hydrolytic step on performance of the digesters, shifts in degradation pathways could only be identified from the activity assay results. The activity assays revealed a shift in the consumption rates of acetogenic substrates as the SRT was lowered. Specifically, the degradation rate of butyrate and propionate decreased with the lowering of the SRT while the ethanol degradation rate followed an inverted trend (higher degradation rate at lower SRT).

Butyrate and ethanol are derivatives of the Acetyl-CoA branch of the EMP glycolysis pathway, and shifts in the fermentation pathways are regulated by kinetic and thermodynamic limitations (Hoelzle et al. 2014, Temudo et al. 2007). From the increase in ethanol activity rate, it could be inferred that fermentation towards ethanol gained dominance at lower SRT, with the shift in fermentation pathway potentially due to: (i) the accumulation of fermentation products and (ii) microbial washout. As recently shown by Junicke et al. (2016), different levels of H2 partial pressure in the liquid redirect the electron fluxes in AD, where at partial pressure in the biogas exceeding 40Pa

(0.039% H2) fermentation products were directed towards more reduced compounds (alcohols) due to higher H2 inhibition thresholds for ethanol than for butyrate. In the present study, the concentration of H2 was not measured in the liquid, and in the biogas, it was below detection limit (0.05%). However, as shown in Zheng et al. (2015), dissolved hydrogen can achieve concentrations up to 3% (6kPa equivalent) in the liquid phase, being undetected in the biogas phase. Propionate was detected in the digesters, being the second most abundant VFA at 15-d SRT, 8-d SRT, and 4-d SRT, and the dominant VFA at 2-d SRT for SL and PL (Figure B-I). However, its degradation rate decreased as the SRT was shortened (Figure 20). Although it is well-known that butyrate, ethanol and propionate degradation are H2-dependant and require low thresholds of H2 to proceed (Batstone et al. 2002, van

Lingen et al. 2016) the role of H2 at the lower SRTs cannot be conclusively resolved.

Another possible explanation for the decreasing butyrate and propionate degradation rates can be related to the washout of microbes capable of VFA degradation. Using literature to assign a probable function to the identified taxa, the relative abundance of microbes capable of butyrate and propionate degradation such as Syntrophus, Syntrophobacteraceae and Syntrophorhabdaceae was strongly correlated to SRT (Pearson  > 0.37, P < 0.006), decreasing from an average relative abundance of 3.9% at 15-d SRT to an average of 1.7% at 2-d SRT.

64 The activity assays also showed a shift in the predominance of the main methanogenic pathways. The HM rate was constant at all SRT, but at 4-d and 2-d SRT the rate of AM decrease to achieve similar or lower rates than HM, indicating a higher relative importance of the HM pathway at the lower SRTs. However, neither the relative abundance of methanogens nor their distribution could explain this shift in the pathway. This is similar to the previous results in Chapter 5 where no direct correlation between activity rates and methanogenic archaea relative abundance could be obtained.

6.4.3 Influence of SRT on microbial communities Microbial community changes were mainly driven by the SRT (Figure 21), where two main clusters could be distinguished, one for the 15 and 8-d SRT and another for the 4 and 2-d SRT (Figure B-VI) indicating that the most significant change happened when the system transitioned from 8-d to 4-d SRT. The change in the microbial community was concomitant to the shift of the activity rates (Figure 19B, Figure 20). The decrease from 8-d SRT to 4-d SRT especially affected the relative abundance of the acetoclastic methanogenic population (Methanosaeta), which decreased from the most abundant taxa to the third most abundant taxa at 4-d SRT in all the digesters. The observed drop in the Methanosaeta dominance could have had a role in regulating the development of the microbial communities since Methanosaeta fulfils the niche of converting acetate to methane, the latter exiting the process as a gas.

The decrease from 4-d SRT to 2-d SRT was not accompanied by a further change in the microbial communities, indicating that the pressure applied to microorganisms was not sufficient to alter the microbial community balance or the product profile. Instead, the formation of biofilms covering the digesters paddles and walls to different extents. The presence and formation of biofilm in all digesters provides evidence that the different microbial communities in the digesters developed similar survival strategies under non-favourable methanogenic conditions as a consequence of the operation at low SRTs. However, further research is needed to evaluate how biofilms microbial communities were assembled, the correlations with the suspended biomass, and the effects (e.g maintenance of methanogenesis) on AD process performance.

Concomitant to the SRT decrease, the pH of the digesters media dropped from 7.3 ± 0.1 at 15-d SRT to 6.5 ± 0.1 at 2-d SRT (see table B-I). Despite the significant correlation between pH and the change in microbial communities observed in the PCA plot (Figure 21), previous studies have shown that small changes in pH around neutrality are not a major driver for complex microbial communities (Latif et al., 2015; Latif et al., 2017; Zhalnina et al., 2015). Moreover, in the present study, the largest change in pH occurred between 15-d SRT and 8-d SRT (from 7.3 ± 0.1 to 6.9 ± 0.1) where the microbial communities, metabolic rates and process performance were highly similar between both

65 periods (Figure 19, Figure 20 and Figure 21). In contrast, the most significant change observed between 8-d SRT to 4-d SRT, only decreased the pH from 6.9 ± 0.1 to 6.7 ± 0.1. Therefore, at the studied range, it is suggested that SRT was a stronger ecological driver than the pH despite the significant mathematical correlation between those operational parameters and the microbial communities.

6.4.4 Effect of an operational disturbance on microbial assembly At 4-d SRT the SL community diverged from SS, PL and BG, probably due to an operational disturbance coinciding with the divergence point (Figure 21). While it cannot be conclusively resolved why several OTUs relative to Ruminococcaceae and an OTU relative Ruminococcus succeeded in SL but diminished in SS, PL and BG; biotic factors (i.e. inter-species interactions) may have played a significant role in shaping the microbial communities triggered by the short starvation period. Ruminococcaceae are well-known for their cellulolytic capacity, and as reviewed in Flint et al. (2008), cellulolytic bacteria are influenced by the presence of hydrogen scavengers. The relative abundance of hydrogenotrophic methanogens was fairlyconstant during the SRTs and between digesters; however, 3 OTUs belonging to the family Methanoregulaceae were more expressed in SL than in the other digesters, suggesting a mutualistic effect between Methanoregulaceae and Ruminococcaceae.

Another possibility may be related to an interaction between hydrolytic bacteria and acetoclastic methanogens. Similar to hydrogen scavengers removing the H2 produced during the fermentation, a faster AM rate may have altered the microbial equilibria or created different niches. As shown in Figure 20 and Figure B-V, at 4-d and 2-d SRT the relative acetate-cellulose rate was higher for SL than for the other 3 digesters suggesting that different metabolic equilibriums were achieved either triggered by a different microbial assembly or triggering the change in community composition. However, from a broad perspective, this did not affect the process yields (methane and VFA), as hydrolysis rates were similar through all SRTs (Figure 20) independently of the hydrolytic populations found in each digester.

66 6.5 Conclusions A sequential decrease in SRT from 15 to 8 to 4 and 2 days while holding the organic loading rate -1 -1 constant at 1 g COD Lr d showed that acetoclastic methanogenesis carried out by Methanosaeta remained active down to 2-d SRT in mesophilic anaerobic digesters. Steady state performance was achieved at all studied SRTs, with hydrolysis the main limiting step at longer SRT and a combination of hydrolysis, secondary fermentation and acetoclastic methanogenesis limiting at short SRT. Therefore, only minor accumulation of VFA was achieved (less than 3.5% influent COD). These results indicate that SRT as an individual selection parameter based on growth rates is not strong enough to drive anaerobic digestion towards VFA production.

Microbial activity assays showed a shift in the acetogenic pathways. The degradation rate of butyrate and propionate decreased with the lowering of the SRT while the ethanol degradation rate followed an inverted trend. Moreover, the results show a higher relative importance of the hydrogenotrophic methanogenic pathway at lower SRTs. These observations highlight the importance to utilise independent kinetic assays to gain a better understanding of the anaerobic digestion pathways.

SRT was found to be the main factor affecting the dynamics of the associated microbial communities driving changes at family or genus level. However, the changes in the microbial relative abundance could not be correlated with the shifts in anaerobic digestion pathways. Biofilm growth was observed on the walls and mixing paddle of all digesters with the decrease in SRT, especially noticeable at 2-d SRT. Biofilm formation is hypothesized as a survival response under non-favourable conditions. Interestingly, Methanosaeta was found in high abundance (11 - 44%) within the biofilm communities in all digesters, suggesting biofilm formation was an important step in maintaining the methanogenic function at short SRTs.

Chapter 7

7 Semi-aerobic fermentation as a novel pre- treatment to obtain VFA and increase methane yield from primary sludge

In: Peces, M., Astals, S., Clarke, W.P. and Jensen, P.D., 2016. Semi-aerobic fermentation as a novel pre-treatment to obtain VFA and increase methane yield from primary sludge. Bioresource technology, 200, pp.631-638.

68 7.1 Introduction Municipal wastewater treatment is a core requirement of urban populations and results in the generation of large amounts of sewage sludge where wastewater pollutants such as organic matter, nutrients, heavy metals, and pathogens are collected and concentrated. Sewage sludge management is a major issue since up to one-half of the costs of operating municipal waste water treatment plants are associated with sludge treatment and disposal (Lens 2004). Sewage sludge is conventionally anaerobically digested to recover energy, but there is a growing trend to use sludge as a feedstock in other value-add processes (Zacharof and Lovitt 2013).

Anaerobic digestion (AD) is a series of biochemical processes where organic matter is converted to biogas by a complex microbial community. The methane content of biogas is an important source of renewable energy, and AD processes are net-energy producing rather than net-energy consuming processes. AD also provides avenues to produce a variety of value-add products such as volatile fatty acids (VFA) (i.e. acetate, propionate, and butyrate). VFA are intermediate products during the AD process. Thus, they are being produced and consumed simultaneously. However, different strategies such as shortening digestion times or decoupling the biochemical reactions involved in the fermentation and methanogenic steps can promote the accumulation of VFA, which can be harvested.

VFA have several potential applications within the wastewater treatment plant (WWTP); for instance, VFA can be used to aid biological nutrient removal, replacing expensive carbon sources, such as methanol (Münch et al. 1999). VFA also have stand-alone value as commodity chemicals or precursors used in the production of renewable plastics and biotextiles (Zacharof and Lovitt 2013). However, the desired VFA profile is highly dependent on its subsequent application, either within the WWTP or the commodity value. Within WWTP applications, the preferred VFA for denitrification is acetate followed by butyrate and propionate (Elefsiniotis and Wareham 2007, Gali et al. 2006), whereas biological phosphorus removal processes typically require an acetate to propionate ratio ranging from 0.25 to 0.75 (Broughton et al. 2008, Yuan et al. 2012). Other bioprocesses, such as the production of bioplastics, require different VFA profiles depending on the desired polymers, for example, acetate and butyrate are preferred for polyhydroxybutyrate (PHB) production, while propionate is required when producing polyhydroxyvalerate (PHV) (Shen et al. 2014).

Fermentation of primary sludge (PS) has been studied previously with yields varying from 0.1-0.4 gVFA g-1VS (Ahn and Speece 2006, Cokgor et al. 2009, Eastman and Ferguson 1981, Ucisik and Henze 2008). Nonetheless, little attention has been paid to the impact of a pre-fermentation step and VFA extraction on down-stream processes, particularly the energy production from AD. This is particularly important when considering the WWTP as an integrated process, since extracting VFA will decrease the amount of organic matter fed to AD, potentially decreasing the energy recovered.

69 There is a need for research to determine if the benefits of VFA production and extraction from primary sludge outweigh the potential loss in methane value. Optimal configurations will be influenced by two main factors: (i) the cost (capital investment and operating expenses) of the extraction process and the revenues obtained from VFA use or sale; and (ii) the impact on methane production (Astals et al. 2015).

7.2 Aim and approach The aim of the present study is to evaluate the impact of a pre-fermentation step on subsequent methane yield from primary sludge. The fermentation conditions considered different temperatures (20, 37, 55, and 70 ºC), fermentation periods (12, 24, 48, and 72 h), and oxygen availability (semi- aerobic or anaerobic conditions). The anaerobic biodegradability after fermentation was evaluated using biochemical methane potential (BMP) tests and mathematical modelling.

In Figure 23 and Table 6 it is summarised the rationale, experimental set-up, and materials and methods to accomplish the aim of the present study. The full description of the materials and methods used in this study is described in Chapter 4.

Figure 23. Schematic representation of the rationale and experimental set-up used in Chapter 7

Table 6. List of materials and methods used in Chapter 7 previously described in Chapter 4 Materials Methods - Primary sludge (PS) - Fermentation batch assays - Digested sewage sludge (SS) - Biochemical methane potential test - Analytical methods: o TS, VS, tCOD, sCOD, pH, VFA, Biogas composition - Data analysis: o AD yields o Overall methane yield o BMP modelling (1st order kinetics)

70 7.3 Results and Discussion

7.3.1 Extraction of valuable compounds from primary sludge 7.3.1.1 Organic matter solubilisation

Figure 24 represents the breakdown of COD during fermentation at all tested conditions. After fermentation at 20 ºC (semi-aerobic and anaerobic) and 37 ºC, the soluble COD in the effluent was completely composed of VFA. In contrast, after thermophilic (55 – 70 ºC) fermentation approximately 65% of soluble COD in the effluent was undetermined soluble substances (e.g. saccharides, amino acids, and long chain fatty acids). In all scenarios, the solubilisation yield increased with the temperature and the treatment time.

A B

C D

E F

Figure 24. COD fractionation (in percentage) for each fermentation condition. (A) 20 ºC Semi- aerobic, (B) 20 ºC Semi-aerobic replicated, (C) 20 ºC Anaerobic, (D) 37 ºC Anaerobic, (E) 55 ºC Anaerobic, and (F) 70 ºC Anaerobic. ( )

71 At 20 ºC, the sCOD increased gradually up to 8% of the initial tCOD (anaerobic, 72 h), while at 70 ºC the sCOD increased more rapidly to 16% of the initial tCOD (anaerobic, 72 h) (Figure 24E). These results indicate that when VFA recovery is the main process objective, fermentation should occur at psychrophilic or mesophilic conditions. Nonetheless, fermentation at 20 ºC and 37 ºC also resulted in 9 and 14% COD losses (anaerobic, 72 h), respectively, due to carbon mineralisation (Table C-I). COD mineralisation was not related to methanogenesis or hydrogen production since these products accounted for less than 1% of the initial COD at 20 ºC (anaerobic) and approximately 1% of the initial COD at 37 ºC after 72 h of fermentation. Therefore, the COD mineralisation was hypothesised to be due to the COD consumption from other processes such as sulphate reduction. COD mineralisation at thermophilic conditions was negligible. This factor should be considered when estimating energy recovery in the WWTP since uncontrolled COD mineralisation during fermentation reduces VFA recovery efficiency as well as the organic matter available for methane production in the subsequent anaerobic digestion step.

7.3.1.2 VFA distribution Controlling fermentation products during mixed culture fermentation of complex substrates is a difficult task with process performance depending on several factors such as substrate composition, temperature, pH, retention time, and the microbial community. Despite this, some VFA distribution trends could be observed depending on the fermentation temperature and treatment time (Figure 25).

Acetate was, regardless the treatment time, the major VFA contributor at 20 ºC (semi-aerobic and anaerobic), 55 ºC and 70 ºC. Among them, fermentation at 20 ºC, either semi-aerobically or -1 anaerobically, delivered the richest acetate stream (1.63, and 2.26 gAcetate L , respectively). PS fermentation at 37 ºC favoured propionate production (Figure 25D), with an acetate to propionate ratio of 0.77 (COD-basis) after 72h fermentation. Propionate was also obtained, in lower amounts, after 20 ºC fermentation (both semi-aerobic and anaerobic); whereas 55ºC fermentation led to the accumulation of butyrate and ethanol (Figure 25E). Similar VFA distribution profiles have been reported by Ahn and Speece (2006) and Ucisik and Henze (2008) when fermenting PS under similar conditions. In contrast to Tsapekos et al. (2017), the VFA distribution was not affected by the oxygen availability. In the current experimental design, oxygen was not mixed within the substrate as opposed to Tsapekos et al. (2017), therefore only the top layer of the sludge was in contact with the air (Figure 27). This stratification suggests that anaerobic fermentation could have been occurring at the bottom of the bottle from where VFA were recovered. Regarding the net VFA production, the highest -1 acidification yield was reached at 37 ºC (143 gCODVFA kg VS) followed by 20 ºC anaerobic -1 -1 (65 gCODVFA kg VS) and 20 ºC semi-aerobic (43 gCODVFA kg VS). The lowest acidification yields

72 -1 were obtained at thermophilic conditions, with 23 gCODVFA kg VS at 55 ºC, and negligible at 70 ºC (Table 7). Therefore, different pre-treatment conditions would be required depending on the amount of VFA required and the desired profile.

A B

C D

E F

Figure 25. VFA distribution depending of (A) 20 ºC Semi-Aerobic, (B) 20 ºC Semi-Aerobic (batch 2), (C) 20 ºC Anaerobic, (D) 37 ºC Anaerobic, (E) 55 ºC Anaerobic, and (F) 70 ºC Anaerobic, at the different exposure times applied. ( )

73 The low acidification yields observed at 55 and 70 ºC could be attributed to the slow development of an anaerobic thermophilic culture during fermentation since no adapted inoculum was used, and the test conditions relied on the activity of native PS microorganisms. However, higher acidogenic activities have been reported at thermophilic temperatures up to and exceeding 70 ºC (Bolzonella et al. 2007, Ge et al. 2011b, Lu et al. 2008), suggesting that different results might be observed when using microbes acclimatised to this temperature range. Finally, the pH at 20 ºC semi-aerobic, 20 ºC anaerobic and 37 ºC anaerobic drifted from 5.10 (0 h) to 4.76, 4.66 and 4.78, respectively (72 h). In contrast, the thermophilic test showed a slight increase in pH after 72 h treatment time. The pH increased to 5.30 at 55 ºC anaerobic and 5.32 at 70 ºC anaerobic.

Table 7. Summary of solubilisation and acidification yields and methane yields for all the fermentation conditions

74 7.3.2 Extraction of soluble compounds and influence on PS methane yield Figure 26 displays the experimental methane production profiles of the solid fraction of pre- fermented PS for all experimental conditions.

A B

C D

E F

Figure 26. Cumulative specific methane production curves after fermentation at different temperatures, exposure time, and control. (A) 20 ºC Semi-Aerobic, (B) 20 ºC Semi-Aerobic (batch 2), (C) 20 ºC Anaerobic, (D) 37 ºC Anaerobic, (E) 55 ºC Anaerobic, and (F) 70 ºC Anaerobic.

75 The results show that the fermentation step under anaerobic conditions neither favoured nor -1 decreased the solid-fraction B0 with values of approximately 340 LCH4 kg VS for all anaerobic fermentation temperatures and treatment times (Table 7), showing no statistical differences between themselves and the control (see Figure C-I). However, 20 ºC semi-aerobic conditions significantly increased B0 compared to the control (primary sludge without pre-fermentation). Specifically, B0 after -1 -1 72h fermentation increased from 336 ± 14 LCH4 kg VS to 382 ± 14 LCH4 kg VS (P < 0.009; two- tail paired t-test ), which represents a methane potential increase of 14%. The second batch of primary sludge pre-fermented at 20 ºC semi-aerobic conditions showed a lower B0 increase (Figure 26-B) compared to the first batch (Table 7). However, the increase was statistically significant (P < 0.015, two-tail paired t-test), indicating that despite the dependency of the increase on the characteristics of the PS, pre-fermentation at 20 ºC semi-aerobic conditions is able to improve the methane yield and recover VFA.

Substrate B0 is a common parameter for assessing the feasibility and expected the performance of AD processes. However, this parameter does not reflect the methane losses due to removal and mineralisation of organic material during waste processing (e.g. pre-treatment, VFA/product recovery, sulphate reduction) (Astals et al. 2015). Therefore, the overall methane yield (B’) was used to evaluate the influence of product extraction and uncontrolled COD losses on the methane production regarding PS initial organic matter content. As shown in Table 7, anaerobic fermentation decreased sludge B’ up to a 21%. Thus, under anaerobic conditions VFA recovery and COD losses, significant at 20 ºC (9%) and 37 ºC (14%), contributed to reducing the methane production in the subsequent AD step. Nevertheless, 20 ºC semi-aerobic fermentation resulted in a minor, but statistically significant increased methane yield even after VFA removal and COD losses (5%) occurring during the fermentation. This phenomenon coincided with the formation of a white mouldy- like layer on the top of the sludge. The layer was not uniform and consisted of round white patches distributed along the sludge surface, being more spread and prominent after 48h treatment time (especially noticeable at 72h and 96h). Considering the test conditions and the physical appearance of the biomass, it is hypothesised that the organisms were fungi (Figure 27).

76 A B

Figure 27. Primary sludge after 72 h of fermentation (20 ºC Semi-Aerobic). (A) Graphic representation. (B) Photograph. 1: liquid rich in VFA; 2: residual sludge; and 3: semi-aerobic layer colonised by fungi.

Several groups of fungi have been found in municipal sewage sludge (Fakhru’l-Razi et al. 2002, Kacprzak et al. 2005). These fungi are versatile organic matter consumers, especially at low pH, where bacterial growth is hindered (More et al. 2010). However, in this study the pH of the sludge varied from 5.1 to 4.8, high enough to sustain acidogenic activity. Therefore, in the semi-aerobic conditions, where fungi were observed, fungi may have partially depolymerised complex structures and made available a greater portion of the PS to the fermentative bacteria. Fungi have the capability to degrade cellulose, hemicellulose and polysaccharides by excreting extracellular enzymes (Pointing 2001) although they are best known for excreting extracellular lignin modifying enzymes that perform lignin degradation. This quality is much less common in anaerobic microorganisms and has prompted the use of fungi as a pre-treatment to enhance the methane potential of lignocellulosic substrates, otherwise difficult to degrade anaerobically (Zheng et al. 2014). In the present study, fungi may have improved the biodegradability of the pre-fermented sludge under semi-aerobic conditions making it more accessible for the subsequent anaerobic microbes; thereby enhancing the overall methane yield (B’).

While the specific role of fungi in this study is not completely elucidated, results were repeatable and confirmed the phenomenon (Table 7). However, the magnitude of the effect in the replicated experiment, regarding acidification yield and methane production, was lower than for the first batch of PS. In either way, results indicate that 20 °C semi-aerobic fermentation was the only configuration that allowed both VFA recovery and an increase methane production, thereby enhancing overall resource recovery. Results also suggest that a similar phenomenon (i.e. VFA recovery and increased

77 methane recovery) could happen at 37 °C and, to a minor degree, 20 °C anaerobic fermentation if the COD mineralisation mechanisms could be minimised (Table C-I and Figure C-I).

7.4 Conclusions Primary sludge fermentation conditions affected solubilisation yields, VFA profile and methane recovery potential. At 20 ºC (both semi-aerobic and anaerobic) and 37 ºC anaerobic, solubilised COD was mainly VFA with acetate and propionate the major contributors. Similar VFA distributions where observed independently of the oxygen availability (semi-aerobic or anaerobic conditions at 20 ºC) suggesting a low influence of those conditions during the pre-fermentation step. At thermophilic temperature, 55 and 70 ºC, solubilised COD was mainly other organic compounds. Anaerobic sludge fermentation (37, 55 and 70 ºC) led to higher solubilisation yields but reduced subsequent methane potential by 20%. However, semi-aerobic fermentation at 20 ºC allowed VFA production (43 -1 gCODVFA kg VS) and a statistically significant improvement in methane potential. The latter phenomenon was linked to fungi observed growing on the top layer of sludge during fermentation.

Chapter 8

8 Conclusions and recommendations

As discussed in the literature review, one of the main challenges for producing higher value organic compounds (as compared to biogas) from anaerobic digestion is the apparent low controllability of product formation and specificity from biodegradable wastes by mix-culture microbial processes. Therefore, the overall aim of this thesis was to gain a fundamental understanding of how complex microbial communities interact and function in the anaerobic digestion process. To achieve this aim, this thesis has integrated well-controlled long-term experiments (bioreactor operation), metabolic rates evaluation (activity assays), and microbial community profiling (16S rRNA gene sequencing).

The present chapter integrates the overall conclusions of this thesis and future directions, whereas the targeted conclusions derived from the experimental work are found at the end of each result chapter.

8.1 Overall conclusions Overall, this thesis has shown that the selective pressures induced by operational conditions influence the development of the composition of microbial communities, directing and dictating their dynamics towards a similar structure independently of the initial species composition.

This is a significant contribution as it expands on the knowledge of previous studies in the microbial ecology field which show that under well-controlled environments, the behaviour of a

79 microbial community can be reproduced by replicating the selection pressures. However, this study has shown that selection pressures can shape different microbial communities progressively driving them towards a core microbiome. When given enough time, the different microbial communities developed similarly under the imposed selection pressure, with several taxa identified to be positively correlated with certain metabolic rates that comprise the anaerobic digestion process.

Under the rationale that selection pressures can drive the microbial communities and their behaviour (at different ecosystem levels), in this thesis, the solid retention time (linked to microbial growth rates in continuously stirred digesters) was used as a selection pressure. It was found that the selection pressure was able to drive the composition of the major microbial groups as well as their functionality to an extent, yet, statistically significant differences remained in the microbial composition and the products yields for the four communities. For instance, while from a practical point of view the four tested communities increased their volatile fatty acids yield at decreased retention times, differences between the yield achieved and the composition varied significantly. Alternatively, while at lower retention times all communities developed a form of biofilm, the extent and composition of the biofilm organisms was not synchronous. In this thesis, while the variations observed in the product spectrum could not be directly attributed to the microbial communities present, the impact of small differences in taxa impact cannot be overruled. These variations reveal that there are other factors that may need to be considered when the chosen selection pressure is not dominant, suggesting that more research should be directed to understand what is the impact of uncontrolled factors (e.g. apparent stochasticity of the system) to develop strategies to manage them when required.

To contextualise these strategies, a wastewater treatment plant has been used as a case scenario to implement the bio-refinery framework, where primary sludge was the selected biodegradable waste to serve as a substrate to produce value-add compounds (volatile fatty acids and biogas). The fermentation of primary sludge under different operational conditions (oxygen availability, temperature, and retention time) was shown to greatly impact the volatile fatty acid and biogas yield, showing that product formation can be effectively controlled in a repeatable manner. However, replication experiments using a different primary sludge showed that while the main trends could be replicated, the magnitude of the response (i.e., volatile fatty acid and methane yield) was different between primary sludge batches.

All these findings suggest that the composition of microbial communities and their functionality can be effectively driven by operational decisions; however, depending on the selection pressures applied other elements (e.g. the physicochemical characteristics of the waste, native microbial

80 communities, or incoming microorganisms) should be considered to further improve the system controllability.

8.2 Future directions The findings presented in this thesis have contributed to understanding some of the mechanisms that define microbial composition and function; however, there are still considerable challenges that can be further explored.

 The use of non-adapted microbial communities The experiment was carried out using initial microbial communities that despite their differences in taxonomy and metabolic rates, were already suited to carry out the anaerobic digestion processes. This limits the assessment on how microbial composition affects the process functionality, as understanding which functions may be developed over time, and which populations are linked to those functions. This can have practical implications to target communities for processes such as bio-augmentation, define bio-markers, or study how the interaction between microbes is reflected in process performance.

 The use of not permanent changes or step-wise changes Questions relating to microbial resistance, resilience, and functional redundancy can be valuable to help understand how microbial communities respond towards a sudden change. To explore this avenue of investigation, operational changes can be switched between two (or more) states; returning to an operational baseline after a change. This, combined with the existing literature, may help to understand further how microbes develop survival strategies or return to previous states.

 The use of real feedstocks (biodegradable waste) Most of the experiments carried out in this thesis have used a synthetic feedstock based on a controlled mixture between a model carbohydrate (cellulose) and a model protein (casein), avoiding the variability caused by the substrate heterogeneity and microbial influx. However, as shown in the last experiment, and in previous literature, feedstock composition may play an important role in defining boundaries of product profile controllability.

81  The development of a standardised methodology for kinetic assessments In this thesis metabolic activity rate assessments have been based on the methodology described by Soto et al., (1993) Methanogenic and non-methanogenic activity tests. Theoretical basis and experimental set-up. Despite the number of comparisons and correlations that have been able to be drawn in this thesis, the results are hardly comparable with other literature because of lack of a unified methodology. Several factors should be considered for the development of a robust method such as: (i) the amount of inoculum required for the test, (ii) the amount of substrate, (iii) the type substrate to be tested, (iv) inoculum background correction, (v) kinetic modelling, and (vi) kinetic rates normalisation.

 The use of post-genomic approaches With the increased access and expansion of genomic approaches, it has become evident that microbial communities harbour extensive genotypic variability. However, it remains inconclusive when this extensive genetic heterogeneity is functionally relevant. The inclusion of post-genomic approaches (i.e. metatranscriptomics, metaproteomics, and metabolomics) integrated with process performance and kinetic (metabolic rates) data can offer a platform to understand the mechanisms of microbiological mediated processes.

References

AFMB, CNRS and d'Aix-Marseille, U. ((last visit: 05/06/2017)) http://www.cazy.org/. Agler, M.T., Werner, J.J., Iten, L.B., Dekker, A., Cotta, M.A., Dien, B.S. and Angenent, L.T. (2012) Shaping reactor microbiomes to produce the fuel precursor n-butyrate from pretreated cellulosic hydrolysates. Environ Sci Technol 46(18), 10229-10238. Agler, M.T., Wrenn, B.A., Zinder, S.H. and Angenent, L.T. (2011) Waste to bioproduct conversion with undefined mixed cultures: the carboxylate platform. Trends Biotechnol 29(2), 70-78. Ahn, J.H. and Forster, C., F (2002) The effect of temperature variations on the performance of mesophilic and thermophilic anaerobic filters treating a simulated papermill wastewater. Process Biochrmistry 37, 589–594. Ahn, Y.H. and Speece, R.E. (2006) Elutriated acid fermentation of municipal primary sludge. Water Res 40(11), 2210-2220. Ahring, B.K. (1994) Status on science and application of thermophilic anaerobic digestion. Water Sci Technol 30(12), 241-249. Allgaier, M., Reddy, A., Park, J.I., Ivanova, N., D'Haeseleer, P., Lowry, S., Sapra, R., Hazen, T.C., Simmons, B.A., VanderGheynst, J.S. and Hugenholtz, P. (2010) Targeted discovery of glycoside hydrolases from a switchgrass-adapted compost community. PLoS One 5(1), e8812. Allison, S.D. and Martiny, J.B. (2008) Colloquium paper: resistance, resilience, and redundancy in microbial communities. Proc Natl Acad Sci U S A 105 Suppl 1, 11512-11519. Angelidaki, I., Alves, M., Bolzonella, D., Borzacconi, L., Campos, J.L., Guwy, A.J., Kalyuzhnyi, S., Jenicek, P. and van Lier, J.B. (2009) Defining the biomethane potential (BMP) of solid organic wastes and energy crops: a proposed protocol for batch assays. Water Sci Technol 59(5), 927-934. Angelidaki, I. and Batstone, D.J. (2011) Solid Waste Technology & Management. Christensen, T.H. (ed), pp. 583-600, Blackwell Publishing Ltd, Chichester, UK.

83 Appels, L., Baeyens, J., Degrève, J. and Dewil, R. (2008) Principles and potential of the anaerobic digestion of waste-activated sludge. Progress in Energy and Combustion Science 34(6), 755-781. Arpigny, J.L. and Jaeger, K.E. (1999) Bacterial lipolytic enzymes: classification and properties. Biochemical Journal 343, 177–183. Astals, S., Musenze, R.S., Bai, X., Tannock, S., Tait, S., Pratt, S. and Jensen, P.D. (2015) Anaerobic co-digestion of pig manure and algae: Impact of intracellular algal products recovery on co-digestion performance. Bioresource Technology 181, 97-104. Barrett, A.J., Rawlings, N.D., Salvesen, G. and Fred Woessner, J. (2013) Handbook of Proteolytic Enzymes, 3rd Edition. Barrett, A.J., Rawlings, N.D. and Fred Woessner, J. (eds), pp. li-liv, Elsevier Ltd, London, UK. Batstone, D.J., J. Keller, I. Angelidaki, S.V. Kalyuzhnyi, S.G. Pavlostathis, A. Rozzi, W.T.M. Sanders, Siegrist, H. and Vavilin, V.A. (2002) Anaerobic Digestion Model No.1, IWA Publishing, Cornwall, UK. Batstone, D.J., Tait, S. and Starrenburg, D. (2009) Estimation of hydrolysis parameters in full-scale anerobic digesters. Biotechnol Bioeng 102(5), 1513-1520. Batstone, D.J. and Virdis, B. (2014) The role of anaerobic digestion in the emerging energy economy. Curr Opin Biotechnol 27, 142-149. Bergmann, G.T., Craine, J.M., Robeson, M.S., 2nd and Fierer, N. (2015) Seasonal Shifts in Diet and Gut Microbiota of the American Bison (Bison bison). PLoS One 10(11), e0142409. Berlemont, R., Allison, S.D., Weihe, C., Lu, Y., Brodie, E.L., Martiny, J.B. and Martiny, A.C. (2014) Cellulolytic potential under environmental changes in microbial communities from grassland litter. Front Microbiol 5, 639. Bialek, K., Cysneiros, D. and O'Flaherty, V. (2014) Hydrolysis, acidification and methanogenesis during low-temperature anaerobic digestion of dilute dairy wastewater in an inverted fluidised bioreactor. Appl Microbiol Biotechnol 98(20), 8737-8750. Bialek, K., Kumar, A., Mahony, T., Lens, P.N. and O'Flaherty, V. (2012) Microbial community structure and dynamics in anaerobic fluidized-bed and granular sludge-bed reactors: influence of operational temperature and reactor configuration. Microb Biotechnol 5(6), 738-752. Bolger, A.M., Lohse, M. and Usadel, B. (2014) Trimmomatic: A flexible trimmer for Illumina sequence data. Bioinformatics 30(15), 2114-2120. Bolzonella, D., Cavinato, C., Fatone, F., Pavan, P. and Cecchi, F. (2012) High rate mesophilic, thermophilic, and temperature phased anaerobic digestion of waste activated sludge: a pilot scale study. Waste Manag 32(6), 1196-1201. Bolzonella, D., Pavan, P., Zanette, M. and Cecchi, F. (2007) Two-Phase Anaerobic Digestion of Waste Activated Sludge: Effect of an Extreme Thermophilic Prefermentation. Industrial and Engineering Chemistry Research 46, 6650-6655. Bourque, J.S., Guiot, S.R. and Tartakovsky, B. (2008) Methane production in an UASB reactor operated under periodic mesophilic-thermophilic conditions. Biotechnol Bioeng 100(6), 1115-1121. Bouskova, A., Dohanyos, M., Schmidt, J.E. and Angelidaki, I. (2005) Strategies for changing temperature from mesophilic to thermophilic conditions in anaerobic CSTR reactors treating sewage sludge. Water Res 39(8), 1481-1488. Broughton, A., Pratt, S. and Shilton, A. (2008) Enhanced biological phosphorus removal for high- strength wastewater with a low rbCOD:P ratio. Bioresour Technol 99(5), 1236-1241.

84 Buchan, A., LeCleir, G.R., Gulvik, C.A. and Gonzalez, J.M. (2014) Master recyclers: features and functions of bacteria associated with phytoplankton blooms. Nat Rev Microbiol 12(10), 686-698. Campanaro, S., Treu, L., Cattani, M., Kougias, P.G., Vendramin, V., Schiavon, S., Tagliapietra, F., Giacomini, A. and Corich, V. (2017) In vitro fermentation of key dietary compounds with rumen fluid: A genome-centric perspective. Science of the Total Environment 584-585, 683-691. Caporaso, J.G., Bittinger, K., Bushman, F.D., DeSantis, T.Z., Andersen, G.L. and Knight, R. (2010a) PyNAST: a flexible tool for aligning sequences to a template alignment. Bioinformatics 26(2), 266- 267. Caporaso, J.G., Kuczynski, J., Stombaugh, J., Bittinger, K., Bushman, F.D., Costello, E.K., Fierer, N., Pena, A.G., Goodrich, J.K. and Gordon, J.I. (2010b) QIIME allows analysis of high-throughput community sequencing data. Nature methods 7(5), 335-336. Carballa, M., Regueiro, L. and Lema, J.M. (2015) Microbial management of anaerobic digestion: exploiting the microbiome-functionality nexus. Curr Opin Biotechnol 33, 103-111. Cardenas, J.P., Quatrini, R. and Holmes, D.S. (2016) Genomic and metagenomic challenges and opportunities for bioleaching: a mini-review. Res Microbiol 167(7), 529-538. Cherubini, F. (2010) The biorefinery concept: Using biomass instead of oil for producing energy and chemicals. Energy Conversion and Management 51(7), 1412-1421. Cirne, D.G., Lehtomaki, A., Bjornsson, L. and Blackall, L.L. (2007) Hydrolysis and microbial community analyses in two-stage anaerobic digestion of energy crops. J Appl Microbiol 103(3), 516- 527. Cokgor, E.U., Oktay, S., Tas, D.O., Zengin, G.E. and Orhon, D. (2009) Influence of pH and temperature on soluble substrate generation with primary sludge fermentation. Bioresour Technol 100(1), 380- 386. Cope, J.L., Hammett, A.J., Kolomiets, E.A., Forrest, A.K., Golub, K.W., Hollister, E.B., DeWitt, T.J., Gentry, T.J., Holtzapple, M.T. and Wilkinson, H.H. (2014) Evaluating the performance of carboxylate platform fermentations across diverse inocula originating as sediments from extreme environments. Bioresour Technol 155, 388-394. Dareioti, M.A., Vavouraki, A.I. and Kornaros, M. (2014) Effect of pH on the anaerobic acidogenesis of agroindustrial wastewaters for maximization of bio-hydrogen production: a lab-scale evaluation using batch tests. Bioresour Technol 162, 218-227. De Francisci, D., Kougias, P.G., Treu, L., Campanaro, S. and Angelidaki, I. (2015) Microbial diversity and dynamicity of biogas reactors due to radical changes of feedstock composition. Bioresour Technol 176, 56-64. de Jonge, N., Moset, V., Moller, H.B. and Nielsen, J.L. (2017) Microbial population dynamics in continuous anaerobic digester systems during start up, stable conditions and recovery after starvation. Bioresour Technol 232, 313-320. De Vrieze, J., Christiaens, M.E.R. and Verstraete, W. (2017) The microbiome as engineering tool: Manufacturing and trading between microorganisms. N Biotechnol. De Vrieze, J., Gildemyn, S., Vilchez-Vargas, R., Jáuregui, R., Pieper, D.H., Verstraete, W. and Boon, N. (2015a) Inoculum selection is crucial to ensure operational stability in anaerobic digestion. Appl Microbiol Biotechnol 99(1), 189-199. De Vrieze, J., Raport, L., Willems, B., Verbrugge, S., Volcke, E., Meers, E., Angenent, L.T. and Boon, N. (2015b) Inoculum selection influences the biochemical methane potential of agro-industrial substrates. Microb Biotechnol 8(5), 776-786.

85 Degnan, P.H. and Ochman, H. (2012) Illumina-based analysis of microbial community diversity. ISME J 6(1), 183-194. Donoso-Bravo, A., Retamal, C., Carballa, M., Ruiz-Filippi, G. and Chamy, R. (2009) Influence of temperature on the hydrolysis, acidogenesis and methanogenesis in mesophilic anaerobic digestion: parameter identification and modeling application. Water Sci Technol 60(1), 9-17. Drake, H.L., Küsel, K. and Matthies, C. (2013) The Prokaryotes. Rosenberg, E. (ed), pp. 3-60, Springer- Verlag Berlin Heidelberg. Dürre, P. (2005) Handbook on clostridia, Taylor & Francis Group. CRC Press, Boca Ranton, Florida, USA. Eastman, J.A. and Ferguson, J.F. (1981) Solubilization of Particulate Organic Carbon during the Acid Phase of Anaerobic Digestion. Water Pollution Control Federation 53(5), 352-366. Edgar, R.C. (2010) Search and clustering orders of magnitude faster than BLAST. Bioinformatics 26(19), 2460-2461. El-Mashad, H.M., Zeeman, G., van Loon, W.K., Bot, G.P. and Lettinga, G. (2004) Effect of temperature and temperature fluctuation on thermophilic anaerobic digestion of cattle manure. Bioresour Technol 95(2), 191-201. Elefsiniotis, P. and Wareham, D.G. (2007) Utilization patterns of volatile fatty acids in the denitrification reaction. Enzyme and Microbial Technology 41(1-2), 92-97. European-Union (1999) Council Directive 1999/31/EC of 26 April 1999 on the landfill of waste. Union, T.C.o.t.E. (ed), p. 19, Official Journal of the European Communities. Fakhru’l-Razi, A., Zahangir Alam, M., Idris, A., Abd-Aziz, S. and Molla, A.H. (2002) Filamentous Fungi in Indah Water Konsortium (Iwk) Sewage Treatment Plant for Biological Treatment of Domestic Wastewater Sludge. Journal of Environmental Science and Health, Part A 37(3), 309-320. Fang, H.H.P. and Liu, H. (2002) Effect of pH on hydrogen production from glucose by a mixed culture. Bioresource Technology 82, 87 - 93. Farag, I.F., Davis, J.P., Youssef, N.H. and Elshahed, M.S. (2014) Global patterns of abundance, diversity and community structure of the Aminicenantes (candidate phylum OP8). PLoS One 9(3), e92139. Ferguson, R.M., Coulon, F. and Villa, R. (2016) Organic loading rate: A promising microbial management tool in anaerobic digestion. Water Res 100, 348-356. Fernández, A., Huang, S., Seston, S., Xing, J., Hickey, R., Criddle, C. and Tiedje, J. (1999) How Stable Is Stable? Function versus Community Composition. Applied and environmental microbiology 65(8), 3697-3704. Flint, H.J., Bayer, E.A., Rincon, M.T., Lamed, R. and White, B.A. (2008) Polysaccharide utilization by gut bacteria: potential for new insights from genomic analysis. Nat Rev Microbiol 6(2), 121-131. Fontana, A., Campanaro, S., Treu, L., Kougias, P.G., Cappa, F., Morelli, L. and Angelidaki, I. (2018) Performance and genome-centric metagenomics of thermophilic single and two-stage anaerobic digesters treating cheese wastes. Water Research 134, 181-191. Forbes, J.D., Knox, N.C., Ronholm, J., Pagotto, F. and Reimer, A. (2017) Metagenomics: The Next Culture-Independent Game Changer. Frontiers in Microbiology 8(1069). Fuerst, J.A. and Sagulenko, E. (2011) Beyond the bacterium: planctomycetes challenge our concepts of microbial structure and function. Nat Rev Microbiol 9(6), 403-413.

86 Gali, A., Dosta, J. and Mata-Alvarez, J. (2006) Use of Hydrolyzed Primary Sludge as Internal Carbon Source for Denitrification in a SBR Treating Reject Water via Nitrite. Industrial and Engineering Chemistry Research 45, 7661-7666. Garcia-Heras, J.L. (2003) Biomethanization of the Organic Fraction of Municipal Solid Wastes. Mata- Alvarez, J. (ed), pp. 21-58, IWA Publishing, Padstow, Cornwall, UK. Gavala, H.N. and Lyberatos, G. (2001) Influence of anaerobic culture acclimation on the degradation kinetics of various substrates. Biotechnol Bioeng 74(3), 181-195. Gavala, H.N., Yenal, U., Skiadas, I.V., Westermann, P. and Ahring, B.K. (2003) Mesophilic and thermophilic anaerobic digestion of primary and secondary sludge. Effect of pre-treatment at elevated temperature. Water Research 37(19), 4561-4572. Ge, H., Jensen, P.D. and Batstone, D.J. (2011a) Relative kinetics of anaerobic digestion under thermophilic and mesophilic conditions. Water Science & Technology 64(4), 848. Ge, H., Jensen, P.D. and Batstone, D.J. (2011b) Temperature phased anaerobic digestion increases apparent hydrolysis rate for waste activated sludge. Water Research 45(4), 1597-1606. Goberna, M., Gadermaier, M., Garcia, C., Wett, B. and Insam, H. (2010) Adaptation of methanogenic communities to the cofermentation of cattle excreta and olive mill wastes at 37 degrees C and 55 degrees C. Appl Environ Microbiol 76(19), 6564-6571. Gomez-Romero, J., Gonzalez-Garcia, A., Chairez, I., Torres, L. and García-Peña, E.I. (2014) Selective adaptation of an anaerobic microbial community: Biohydrogen production by co-digestion of cheese whey and vegetables fruit waste. International Journal of Hydrogen Energy 39(24), 12541-12550. Griffin, J.S. and Wells, G.F. (2017) Regional synchrony in full-scale activated sludge bioreactors due to deterministic microbial community assembly. ISME J 11(2), 500-511. Gu, Y., Chen, X., Liu, Z., Zhou, X. and Zhang, Y. (2014) Effect of inoculum sources on the anaerobic digestion of rice straw. Bioresour Technol 158, 149-155. Gullert, S., Fischer, M.A., Turaev, D., Noebauer, B., Ilmberger, N., Wemheuer, B., Alawi, M., Rattei, T., Daniel, R., Schmitz, R.A., Grundhoff, A. and Streit, W.R. (2016) Deep metagenome and metatranscriptome analyses of microbial communities affiliated with an industrial biogas fermenter, a cow rumen, and elephant feces reveal major differences in carbohydrate hydrolysis strategies. Biotechnol Biofuels 9, 121. Handelsman, J. (2007) Encyclopedia of life sciences (eLS), John Wiley & Sons, Ltd, Chichester, UK. Hanreich, A., Schimpf, U., Zakrzewski, M., Schlüter, A., Benndorf, D., Heyer, R., Rapp, E., Pühler, A., Reichl, U. and Klocke, M. (2013) Metagenome and metaproteome analyses of microbial communities in mesophilic biogas-producing anaerobic batch fermentations indicate concerted plant carbohydrate degradation. Systematic and Applied Microbiology 36(5), 330-338. Hao, L., Lü, F., Mazéas, L., Desmond-Le Quéméner, E., Madigou, C., Guenne, A., Shao, L., Bouchez, T. and He, P. (2015) Stable isotope probing of acetate fed anaerobic batch incubations shows a partial resistance of acetoclastic methanogenesis catalyzed by Methanosarcina to sudden increase of ammonia level. Water Research 69, 90-99. Heyer, R., Benndorf, D., Kohrs, F., De Vrieze, J., Boon, N., Hoffmann, M., Rapp, E., Schluter, A., Sczyrba, A. and Reichl, U. (2016) Proteotyping of biogas plant microbiomes separates biogas plants according to process temperature and reactor type. Biotechnol Biofuels 9, 155. Ho, D., Jensen, P. and Batstone, D. (2014) Effects of temperature and hydraulic retention time on acetotrophic pathways and performance in high-rate sludge digestion. Environ Sci Technol 48(11), 6468-6476.

87 Ho, D.P., Jensen, P.D. and Batstone, D.J. (2013) Methanosarcinaceae and acetate-oxidizing pathways dominate in high-rate thermophilic anaerobic digestion of waste-activated sludge. Appl Environ Microbiol 79(20), 6491-6500. Hoelzle, R.D., Virdis, B. and Batstone, D.J. (2014) Regulation mechanisms in mixed and pure culture microbial fermentation. Biotechnol Bioeng 111(11), 2139-2154. Hoornweg, D. and Bhada-Tata, P. (2012) What a Waste - A Global Review of Solid Waste Management, Washington, DC. USA. Horiuchi, J.I., Shimizu, T., Tada, K., Kanno, T. and Kobayashi, M. (2002) Selective production of organic acids in anaerobic acid reactor by pH control. Bioresource Technology 82, 209-213. Hugenholtz, P. and Tyson, G.W. (2008) Metagenomics. Nature 455(7212), 481-483. Hugenholtz, P., Tyson, G.W. and Blackall, L.L. (2002) Gene probes, pp. 29-42, Springer. Imelfort, M., Parks, D., Woodcroft, B.J., Dennis, P., Hugenholtz, P. and Tyson, G.W. (2014) GroopM: an automated tool for the recovery of population genomes from related metagenomes. PeerJ 2, e409v401. Ito, T., Yoshiguchi, K., Ariesyady, H.D. and Okabe, S. (2011) Identification of a novel acetate-utilizing bacterium belonging to Synergistes group 4 in anaerobic digester sludge. ISME J 5(12), 1844-1856. Jagadabhi, P.S., Kaparaju, P. and Rintala, J. (2010) Effect of micro-aeration and leachate replacement on COD solubilization and VFA production during mono-digestion of grass-silage in one-stage leach- bed reactors. Bioresour Technol 101(8), 2818-2824. Jain, M., Olsen, H.E., Paten, B. and Akeson, M. (2016) The Oxford Nanopore MinION: delivery of nanopore sequencing to the genomics community. Genome Biol 17(1), 239. Jensen, P.D., Ge, H. and Batstone, D.J. (2011) Assessing the role of biochemical methane potential tests in determining anaerobic degradability rate and extent. Water Science & Technology 64(4), 880. Junicke, H., van Loosdrecht, M.C.M. and Kleerebezem, R. (2016) Kinetic and thermodynamic control of butyrate conversion in non-defined methanogenic communities. Appl Microbiol Biotechnol 100(2), 915-925. Kacprzak, M., Neczaj, E. and Okoniewska, E. (2005) The comparative mycological analysis of wastewater and sewage sludges from selected wastewater treatment plants. Desalination 185(1-3), 363-370. Kalmykova, Y., Sadagopan, M. and Rosado, L. (2018) Circular economy – From review of theories and practices to development of implementation tools. Resources, Conservation and Recycling 135, 190-201. Karakashev, D., Batstone, D.J., Trably, E. and Angelidaki, I. (2006) Acetate oxidation is the dominant methanogenic pathway from acetate in the absence of Methanosaetaceae. Appl Environ Microbiol 72(7), 5138-5141. Kent, A.D., Yannarell, A.C., Rusak, J.A., Triplett, E.W. and McMahon, K.D. (2007) Synchrony in aquatic microbial community dynamics. ISME J 1(1), 38-47. Kim, J., Lee, S. and Lee, C. (2013) Comparative study of changes in reaction profile and microbial community structure in two anaerobic repeated-batch reactors started up with different seed sludges. Bioresour Technol 129, 495-505. Kim, M., Ahn, Y.H. and Speece, R.E. (2002) Comparative process stability and efficiency of anaerobic digestion; mesophilic vs. thermophilic. Water Research 36(17), 4369-4385.

88 Kim, M., Gomec, C.Y., Ahn, Y. and Speece, R.E. (2003) Hydrolysis and acidogenesis of particulate organic material in mesophilic and thermophilic anaerobic digestion. Environ Technol 24(9), 1183- 1190. Kircher, M. (2014) The Emerging Bioeconomy: Industrial Drivers, Global Impact, and International Strategies. Industrial Biotechnology 10(1), 11-18. Kirkegaard, R.H., Dueholm, M.S., McIlroy, S.J., Nierychlo, M., Karst, S.M., Albertsen, M. and Nielsen, P.H. (2016) Genomic insights into members of the candidate phylum Hyd24-12 common in mesophilic anaerobic digesters. ISME Journal 10(10), 2352-2364. Krakat, N., Schmidt, S. and Scherer, P. (2011) Potential impact of process parameters upon the bacterial diversity in the mesophilic anaerobic digestion of beet silage. Bioresour Technol 102(10), 5692-5701. Kuever, J. (2014) The Family Syntrophobacteraceae. 289-299. Langenheder, S., Lindstrom, E.S. and Tranvik, L.J. (2006) Structure and function of bacterial communities emerging from different sources under identical conditions. Appl Environ Microbiol 72(1), 212-220. Latif, M.A., Mehta, C.M. and Batstone, D.J. (2015) Low pH anaerobic digestion of waste activated sludge for enhanced phosphorous release. Water Res 81, 288-293. Latif, M.A., Mehta, C.M. and Batstone, D.J. (2017) Influence of low pH on continuous anaerobic digestion of waste activated sludge. Water Res 113, 42-49. Lee, K.C.Y., Dunfield, P.F. and Stott, M.B. (2014) The Phylum Armatimonadetes. 447-458. Lee, S.H., Kang, H.J., Lee, Y.H., Lee, T.J., Han, K., Choi, Y. and Park, H.D. (2012) Monitoring bacterial community structure and variability in time scale in full-scale anaerobic digesters. J Environ Monit 14(7), 1893-1905. Legendre, P. and Legendre, L. (2012) Numerical Ecology, Elsevier, Oxford, UK. Lens, P. (2004) Resource recovery and reuse in organic solid waste management, IWA Publishing, London, UK. Li, Y.-F., Nelson, M.C., Chen, P.-H., Graf, J., Li, Y. and Yu, Z. (2015) Comparison of the microbial communities in solid-state anaerobic digestion (SS-AD) reactors operated at mesophilic and thermophilic temperatures. Appl Microbiol Biotechnol 99(2), 969-980. Lin, Q., De Vrieze, J., Li, C., Li, J., Li, J., Yao, M., Hedenec, P., Li, H., Li, T., Rui, J., Frouz, J. and Li, X. (2017) Temperature regulates deterministic processes and the succession of microbial interactions in anaerobic digestion process. Water Res 123, 134-143. Liu, T., Sun, L., Muller, B. and Schnurer, A. (2017) Importance of inoculum source and initial community structure for biogas production from agricultural substrates. Bioresour Technol 245(Pt A), 768-777. Lombard, V., Golaconda Ramulu, H., Drula, E., Coutinho, P.M. and Henrissat, B. (2014) The carbohydrate-active enzymes database (CAZy) in 2013. Nucleic Acids Res 42(Database issue), D490-495. Louis, P., Hold, G.L. and Flint, H.J. (2014) The gut microbiota, bacterial metabolites and colorectal cancer. Nat Rev Microbiol 12(10), 661-672. Lu, J., Gavala, H.N., Skiadas, I.V., Mladenovska, Z. and Ahring, B.K. (2008) Improving anaerobic sewage sludge digestion by implementation of a hyper-thermophilic prehydrolysis step. J Environ Manage 88(4), 881-889.

89 Lu, Y., Slater, F.R., Mohd-Zaki, Z., Pratt, S. and Batstone, D.J. (2011) Impact of operating history on mixed culture fermentation microbial ecology and product mixture. Water Science & Technology 64(3), 760. Lucas, R., Kuchenbuch, A., Fetzer, I., Harms, H. and Kleinsteuber, S. (2015) Long-term monitoring reveals stable and remarkably similar microbial communities in parallel full-scale biogas reactors digesting energy crops. FEMS Microbiol Ecol 91(3), fiv004-fiv004. Luo, G., De Francisci, D., Kougias, P.G., Laura, T., Zhu, X. and Angelidaki, I. (2015) New steady- state microbial community compositions and process performances in biogas reactors induced by temperature disturbances. Biotechnology for Biofuels 8(1), In Press. Lynd, L.R., Weimer, P.J., van Zyl, W.H. and Pretorius, I.S. (2002) Microbial Cellulose Utilization: Fundamentals and Biotechnology. Microbiology and Molecular Biology Reviews 66(3), 506-577. Madigan, M.T., Martinko, J.M., Stahl, D. and P. Clark, D. (2010) Brock Biology of Microorganisms, Benjamin Cummings, San Francisco, CA. USA. Martiny, J.B., Jones, S.E., Lennon, J.T. and Martiny, A.C. (2015) Microbiomes in light of traits: A phylogenetic perspective. Science 350(6261), aac9323. Marzorati, M., Wittebolle, L., Boon, N., Daffonchio, D. and Verstraete, W. (2008) How to get more out of molecular fingerprints: practical tools for microbial ecology. Environ Microbiol 10(6), 1571- 1581. Maspolim, Y., Zhou, Y., Guo, C., Xiao, K. and Ng, W.J. (2014) Comparison of single-stage and two- phase anaerobic sludge digestion systems - Performance and microbial community dynamics. Chemosphere. Mata-Alvarez, J., Dosta, J., Romero-Güiza, M.S., Fonoll, X., Peces, M. and Astals, S. (2014) A critical review on anaerobic co-digestion achievements between 2010 and 2013. Renewable and Sustainable Energy Reviews 36, 412-427. Mata-Alvarez, J., Macé, S. and Llabrés, P. (2000) Anaerobic digestion of organic solid wastes. An overview of research achievements and perspectives. Bioresource Technology 74, 3-16. Maus, I., Cibis, K.G., Bremges, A., Stolze, Y., Wibberg, D., Tomazetto, G., Blom, J., Sczyrba, A., König, H., Pühler, A. and Schlüter, A. (2016) Genomic characterization of Defluviitoga tunisiensis L3, a key hydrolytic bacterium in a thermophilic biogas plant and its abundance as determined by metagenome fragment recruitment. Journal of Biotechnology 232, 50-60. McIlroy, S.J., Kirkegaard, R.H., Dueholm, M.S., Fernando, E., Karst, S.M., Albertsen, M. and Nielsen, P.H. (2017) Culture-independent analyses reveal novel anaerolineaceae as abundant primary fermenters in anaerobic digesters treating waste activated sludge. Frontiers in Microbiology 8 McMurdie, P.J. and Holmes, S. (2013) Phyloseq: An R Package for Reproducible Interactive Analysis and Graphics of Microbiome Census Data. PLoS One 8(4). More, T.T., Yan, S., Tyagi, R.D. and Surampalli, R.Y. (2010) Potential use of filamentous fungi for wastewater sludge treatment. Bioresour Technol 101(20), 7691-7700. Morgan, X.C. and Huttenhower, C. (2014) Meta'omic analytic techniques for studying the intestinal microbiome. Gastroenterology 146(6), 1437-1448. Mosbaek, F., Kjeldal, H., Mulat, D.G., Albertsen, M., Ward, A.J., Feilberg, A. and Nielsen, J.L. (2016) Identification of syntrophic acetate-oxidizing bacteria in anaerobic digesters by combined protein- based stable isotope probing and metagenomics. ISME J 10(10), 2405-2418. Müller, V. (2008) Encyclopedia of Life Sciences (ELS). , John Wiley & Sons, Ltd, Chichester, UK.

90 Mullings, R. and Parish, J.H. (1984) Mesophilic aerobic Gram negative cellulose degrading bacteria from aquatic habitats and soils. Journal of Applied Bacteriology 57(3), 455-468. Münch, E.V., Keller, J., Lant, P.A. and Newell, R. (1999) Mathematical modelling of prefermenters, Model development and verification. Water Research 33(12), 2757–2768. Murali, N., Srinivas, K. and Ahring, B.K. (2017) Biochemical Production and Separation of Carboxylic Acids for Biorefinery Applications. Fermentation 3(2), 22. Neilson, J.W., Jordan, F.L. and Maier, R.M. (2013) Analysis of artifacts suggests DGGE should not be used for quantitative diversity analysis. J Microbiol Methods 92(3), 256-263. Nelson, W.C. and Stegen, J.C. (2015) The reduced genomes of Parcubacteria (OD1) contain signatures of a symbiotic lifestyle. Front Microbiol 6, 713. Nges, I.A. and Liu, J. (2009) Effects of anaerobic pre-treatment on the degradation of dewatered- sewage sludge. Renewable Energy 34(7), 1795-1800. Nges, I.A. and Liu, J. (2010) Effects of solid retention time on anaerobic digestion of dewatered-sewage sludge in mesophilic and thermophilic conditions. Renewable Energy 35(10), 2200-2206. OCDE (2009) The bioeconomy to 2030: designing a policy agenda, OECD publications. Oksanen, J.B., F.G.; Friendly, M.; Kindt, R.; Legendre, P.; McGlinn, D.; Minchin, P.R.; O'Hara, R.B.; Simpson, G.L.; Solymos, P.; Henry, M.; Stevens, H.; Szoecs, E.; Wagner, H. (2016) Vegan: Community Ecology Package. R package version 2.4-0. Palatsi, J., Laureni, M., Andres, M.V., Flotats, X., Nielsen, H.B. and Angelidaki, I. (2009) Strategies for recovering inhibition caused by long chain fatty acids on anaerobic thermophilic biogas reactors. Bioresour Technol 100(20), 4588-4596. Palatsi, J., Vinas, M., Guivernau, M., Fernandez, B. and Flotats, X. (2011) Anaerobic digestion of slaughterhouse waste: main process limitations and microbial community interactions. Bioresour Technol 102(3), 2219-2227. Peces, M., Astals, S. and Mata-Alvarez, J. (2013) Response of a sewage sludge mesophilic anaerobic digester to short and long-term thermophilic temperature fluctuations. Chemical Engineering Journal 233, 109-116. Peck, M.W., Skilton, J.M., Hawkes, F.R. and Hawkes, D.L. (1985) Effects of temperature shock treatments on the stability of anaerobic digesters operated on separated cattle slurry. Water Research 20(4), 453-462. Perrotta, A.R., Kumaraswamy, R., Bastidas-Oyanedel, J.R., Alm, E.J. and Rodriguez, J. (2017) Inoculum composition determines microbial community and function in an anaerobic sequential batch reactor. PLoS One 12(2), e0171369. Pervin, H.M., Dennis, P.G., Lim, H.J., Tyson, G.W., Batstone, D.J. and Bond, P.L. (2013) Drivers of microbial community composition in mesophilic and thermophilic temperature-phased anaerobic digestion pre-treatment reactors. Water Res 47(19), 7098-7108. Pinto, A.J. and Raskin, L. (2012) PCR biases distort bacterial and archaeal community structure in pyrosequencing datasets. PLoS One 7(8), e43093. Pointing, S. (2001) Feasibility of bioremediation by white-rot fungi. Appl Microbiol Biotechnol 57(1- 2), 20-33. Porter, M.D., Andrus, J.M., Bartolerio, N.A., Rodriguez, L.F., Zhang, Y., Zilles, J.L. and Kent, A.D. (2015) Seasonal Patterns in Microbial Community Composition in Denitrifying Bioreactors Treating Subsurface Agricultural Drainage. Microbial Ecology 70(3), 710-723.

91 Poszytek, K., Pyzik, A., Sobczak, A., Lipinski, L., Sklodowska, A. and Drewniak, L. (2017) The effect of the source of microorganisms on adaptation of hydrolytic consortia dedicated to anaerobic digestion of maize silage. Anaerobe 46, 46-55. Puyol, D., Batstone, D.J., Hulsen, T., Astals, S., Peces, M. and Kromer, J.O. (2016) Resource Recovery from Wastewater by Biological Technologies: Opportunities, Challenges, and Prospects. Front Microbiol 7, 2106. Ramette, A. (2007) Multivariate analyses in microbial ecology. FEMS Microbiol Ecol 62(2), 142-160. Ramsay, I.R. and Pullammanappallil, P.C. (2001) Protein degradation during anaerobic wastewater treatment: derivation of stoichiometry. Biodegradation 12, 247-257. Regueiro, L., Lema, J.M. and Carballa, M. (2015) Key microbial communities steering the functioning of anaerobic digesters during hydraulic and organic overloading shocks. Bioresour Technol 197, 208- 216. Regueiro, L., Veiga, P., Figueroa, M., Alonso-Gutierrez, J., Stams, A.J., Lema, J.M. and Carballa, M. (2012) Relationship between microbial activity and microbial community structure in six full-scale anaerobic digesters. Microbiol Res 167(10), 581-589. Regueiro, L., Veiga, P., Figueroa, M., Lema, J.M. and Carballa, M. (2014) Influence of transitional states on the microbial ecology of anaerobic digesters treating solid wastes. Applied microbiology and biotechnology, 98(5), pp.2015-2027 Rotaru, A.-E., Shrestha, P.M., Liu, F., Shrestha, M., Shrestha, D., Embree, M., Zengler, K., Wardman, C., Nevin, K.P. and Lovley, D.R. (2014) A new model for electron flow during anaerobic digestion: direct interspecies electron transfer to Methanosaeta for the reduction of carbon dioxide to methane. Energy Environ. Sci. 7(1), 408-415. Ruiz-Sánchez, J., Campanaro, S., Guivernau, M., Fernández, B. and Prenafeta-Boldú, F.X. (2018) Effect of ammonia on the active microbiome and metagenome from stable full-scale digesters. Bioresource Technology 250, 513-522. Sanders, W.T.M. (2001) Anaerobic hydrolysis during digestion of complex substrates, Wageningen Universiteit. Sans, C., Mata-Alvarez, J., Cecchi, F., Pavan, P. and Bassetti, A. (1995) Acidogenic fermentation of organic urban wastes in a plug-flow reactor under thermophilic conditions. Bioresource Technology 54(2), 105-110. Schnürer, A. and Jarvis, A. (2009) Microbiological Handbook for Biogas Plants, Avfall Sverige Development, Malmö, Sweden. Shen, L., Hu, H., Ji, H., Cai, J., He, N., Li, Q. and Wang, Y. (2014) Production of poly(hydroxybutyrate- hydroxyvalerate) from waste organics by the two-stage process: focus on the intermediate volatile fatty acids. Bioresour Technol 166, 194-200. Siegrist, H., Vogt, D., Garcia-Heras, J.L. and Gujer, W. (2002) Mathematical model for meso-and thermophilic anaerobic sewage sludge digestion. Environ Sci Technol 36(5), 1113-1123. Smith, A.L., Shimada, T. and Raskin, L. (2017) A comparative evaluation of community structure in full-scale digesters indicates that two-phase digesters exhibit greater microbial diversity than single- phase digesters. Environ. Sci.: Water Res. Technol. 3(2), 304-311. Song, Y.C., Kwon, S.J. and Woo, J.H. (2004) Mesophilic and thermophilic temperature co-phase anaerobic digestion compared with single-stage mesophilic- and thermophilic digestion of sewage sludge. Water Res 38(7), 1653-1662.

92 Soto, M., Méndez, R. and Lema, J.M. (1993) Methanogenic and non-methanogenic activity tests. Theoretical basis and experimental set up. Water Research 27(8), 1361-1376. Speece, R.E. and Kem, J.A. (1970) The effect of short-term temperature variations on methane production. Water Pollution Control Federation 42(11), 1990-1997. St-Pierre, B. and Wright, A.D. (2014) Comparative metagenomic analysis of bacterial populations in three full-scale mesophilic anaerobic manure digesters. Appl Microbiol Biotechnol 98(6), 2709-2717. Stewart, F.J. (2013) Preparation of microbial community cDNA for metatranscriptomic analysis in marine plankton. Methods Enzymol 531, 187-218. Suen, G., Weimer, P.J., Stevenson, D.M., Aylward, F.O., Boyum, J., Deneke, J., Drinkwater, C., Ivanova, N.N., Mikhailova, N., Chertkov, O., Goodwin, L.A., Currie, C.R., Mead, D. and Brumm, P.J. (2011) The complete genome sequence of Fibrobacter succinogenes S85 reveals a cellulolytic and metabolic specialist. PLoS One 6(4), e18814. Sun, L., Toyonaga, M., Ohashi, A., Matsuura, N., Tourlousse, D.M., Meng, X.Y., Tamaki, H., Hanada, S., Cruz, R., Yamaguchi, T. and Sekiguchi, Y. (2015) Isolation and characterization of Flexilinea flocculi gen. nov., sp. nov., a filamentous anaerobic bacterium belonging to the class Anaerolineae in the phylum Chloroflexi. Int J Syst Evol Microbiol. Sun, L., Toyonaga, M., Ohashi, A., Tourlousse, D.M., Matsuura, N., Meng, X.Y., Tamaki, H., Hanada, S., Cruz, R., Yamaguchi, T. and Sekiguchi, Y. (2016) Lentimicrobium saccharophilum gen. nov., sp. nov., a strictly anaerobic bacterium representing a new family in the phylum Bacteroidetes, and proposal of Lentimicrobiaceae fam. nov. Int J Syst Evol Microbiol. Sundberg, C., Al-Soud, W.A., Larsson, M., Alm, E., Yekta, S.S., Svensson, B.H., Sorensen, S.J. and Karlsson, A. (2013) 454 pyrosequencing analyses of bacterial and archaeal richness in 21 full-scale biogas digesters. FEMS Microbiol Ecol 85(3), 612-626. Temudo, M.F., Kleerebezem, R. and van Loosdrecht, M. (2007) Influence of the pH on (open) mixed culture fermentation of glucose: a chemostat study. Biotechnol Bioeng 98(1), 69-79. Tomazetto, G., Hahnke, S., Koeck, D.E., Wibberg, D., Maus, I., Pühler, A., Klocke, M. and Schlüter, A. (2016) Complete genome analysis of Clostridium bornimense strain M2/40T: A new acidogenic Clostridium species isolated from a mesophilic two-phase laboratory-scale biogas reactor. Journal of Biotechnology 232, 38-49. Treu, L., Campanaro, S., Kougias, P.G., Zhu, X. and Angelidaki, I. (2016a) Untangling the Effect of Fatty Acid Addition at Species Level Revealed Different Transcriptional Responses of the Biogas Microbial Community Members. Environmental Science and Technology 50(11), 6079-6090. Treu, L., Kougias, P.G., Campanaro, S., Bassani, I. and Angelidaki, I. (2016b) Deeper insight into the structure of the anaerobic digestion microbial community; The biogas microbiome database is expanded with 157 new genomes. Bioresource Technology 216, 260-266. Tsapekos, P., Kougias, P.G., Vasileiou, S.A., Lyberatos, G. and Angelidaki, I. (2017) Effect of micro- aeration and inoculum type on the biodegradation of lignocellulosic substrate. Bioresour Technol 225, 246-253. Tuan, N.N., Chang, Y.C., Yu, C.P. and Huang, S.L. (2014) Multiple approaches to characterize the microbial community in a thermophilic anaerobic digester running on swine manure: a case study. Microbiol Res 169(9-10), 717-724. Ucisik, A.S. and Henze, M. (2008) Biological hydrolysis and acidification of sludge under anaerobic conditions: the effect of sludge type and origin on the production and composition of volatile fatty acids. Water Res 42(14), 3729-3738.

93 Van Lier, J.B., Rebac, S. and Lettinga, G. (1997) High-rate anaerobic wastewater treatment under psychrophilic and thermophilic conditions. Water Sci Technol 35(10), 199-206. van Lingen, H.J., Plugge, C.M., Fadel, J.G., Kebreab, E., Bannink, A. and Dijkstra, J. (2016) Thermodynamic Driving Force of Hydrogen on Rumen Microbial Metabolism: A Theoretical Investigation. PLoS One 11(10), e0161362. Vanwonterghem, I., Jensen, P.D., Dennis, P.G., Hugenholtz, P., Rabaey, K. and Tyson, G.W. (2014a) Deterministic processes guide long-term synchronised population dynamics in replicate anaerobic digesters. ISME J 8(10), 2015-2028. Vanwonterghem, I., Jensen, P.D., Ho, D.P., Batstone, D.J. and Tyson, G.W. (2014b) Linking microbial community structure, interactions and function in anaerobic digesters using new molecular techniques. Curr Opin Biotechnol 27, 55-64. Vanwonterghem, I., Jensen, P.D., Rabaey, K. and Tyson, G.W. (2015) Temperature and solids retention time control microbial population dynamics and volatile fatty acid production in replicated anaerobic digesters. Sci. Rep. 5. Vanwonterghem, I., Jensen, P.D., Rabaey, K. and Tyson, G.W. (2016) Genome-centric resolution of microbial diversity, metabolism and interactions in anaerobic digestion. Environ Microbiol 18(9), 3144-3158. Vavilin, V.A., Fernandez, B., Palatsi, J. and Flotats, X. (2008) Hydrolysis kinetics in anaerobic degradation of particulate organic material: an overview. Waste Manag 28(6), 939-951. Venkiteshwaran, K., Milferstedt, K., Hamelin, J., Fujimoto, M., Johnson, M. and Zitomer, D.H. (2017) Correlating methane production to microbiota in anaerobic digesters fed synthetic wastewater. Water Res 110, 161-169. Visser, A., Gao, Y. and Lettinga, G. (1993) Effects of short-term temperature increases on the mesophilic anaerobic breakdown of sulfate containing synthetic wastewater. Water Research 27(4), 541-550. Wagner, M., Amann, R., Lemmer, H. and Schleifer, K.H. (1993) Probing activated sludge with oligonucleotides specific for proteobacteria: inadequacy of culture-dependent methods for describing microbial community structure. Applied and environmental microbiology 59(5), 1520-1525. Werner, J.J., Knights, D., Garcia, M.L., Scalfone, N.B., Smith, S., Yarasheski, K., Cummings, T.A., Beers, A.R., Knight, R. and Angenent, L.A. (2011a) Bacterial community structures are unique and resilient in full-scale bioenergy systems. Proceedings of the National Academy of Sciences 108(10), 4158–4163. Werner, J.J., Knights, D., Garcia, M.L., Scalfone, N.B., Smith, S., Yarasheski, K., Cummings, T.A., Beers, A.R., Knight, R. and Angenent, L.T. (2011b) Bacterial community structures are unique and resilient in full-scale bioenergy systems. Proceedings of the National Academy of Sciences of the United States of America 108(10), 4158-4163. Whitman, W.B., Bowen, T.L. and Boone, D.R. (2006) The Prokaryotes, pp. 165-207, Springer-Verlag, New York, USA. Widder, S., Allen, R.J., Pfeiffer, T., Curtis, T.P., Wiuf, C., Sloan, W.T., Cordero, O.X., Brown, S.P., Momeni, B., Shou, W., Kettle, H., Flint, H.J., Haas, A.F., Laroche, B., Kreft, J.-U., Rainey, P.B., Freilich, S., Schuster, S., Milferstedt, K., van der Meer, J.R., Grokopf, T., Huisman, J., Free, A., Picioreanu, C., Quince, C., Klapper, I., Labarthe, S., Smets, B.F., Wang, H., Isaac Newton Institute, F. and Soyer, O.S. (2016) Challenges in microbial ecology: building predictive understanding of community function and dynamics. ISME J 10(11), 2557-2568. WTSI ((last visit: 05/06/2017)) MEROPS - the Peptidase database.

94 Yuan, Z., Pratt, S. and Batstone, D.J. (2012) Phosphorus recovery from wastewater through microbial processes. Curr Opin Biotechnol 23(6), 878-883. Zacharof, M.-P. and Lovitt, R.W. (2013) Complex Effluent Streams as a Potential Source of Volatile Fatty Acids. Waste and Biomass Valorization 4(3), 557-581. Zakrzewski, M., Goesmann, A., Jaenicke, S., Junemann, S., Eikmeyer, F., Szczepanowski, R., Al-Soud, W.A., Sorensen, S., Puhler, A. and Schluter, A. (2012) Profiling of the metabolically active community from a production-scale biogas plant by means of high-throughput metatranscriptome sequencing. J Biotechnol 158(4), 248-258. Zampieri, M., Sekar, K., Zamboni, N. and Sauer, U. (2017) Frontiers of high-throughput metabolomics. Curr Opin Chem Biol 36, 15-23. Zhalnina, K., Dias, R., de Quadros, P.D., Davis-Richardson, A., Camargo, F.A., Clark, I.M., McGrath, S.P., Hirsch, P.R. and Triplett, E.W. (2015) Soil pH determines microbial diversity and composition in the park grass experiment. Microbial ecology 69(2), 395-406. Zheng, H., Zeng, R.J., Duke, M.C., O'Sullivan, C.A. and Clarke, W.P. (2015) Changes in glucose fermentation pathways by an enriched bacterial culture in response to regulated dissolved H2 concentrations. Biotechnol Bioeng 112(6), 1177-1186. Zheng, Y., Zhao, J., Xu, F. and Li, Y. (2014) Pretreatment of lignocellulosic biomass for enhanced biogas production. Progress in Energy and Combustion Science 42, 35-53. Zoetemeyer, R.J., van den Heuvel, J.C. and Cohen, A. (1982) pH influence on acidogenic dissimilation of glucose in an anaerobic digestor. Water Research 16(3), 303-311.

Appendix A

Appendix A: Supplementary material for Chapter 5

96 A.1 Digesters’ performance

A.1.1 Individual VFA A C

B D

Figure A-I. Individual VFA concentration profiles of the 4 digesters (♦) SL, (♦) SS, (♦) PL, (♦) BG. (A) Acetic Acid, (B) Propionate acid, (C) Butyric acid, (D) Valeric acid. Operational stages correspond to the periods (■) Start-up, (■) Transition, and (■) Steady. Data within dotted red lines has been excluded from the data analysis due to operational complications

97 A.1.2 pH, tCOD, TS, and VS A C

B D

Figure A-II. Monitoring parameters profiles of the 4 digesters (♦) SL, (♦) SS, (♦) PL, (♦)BG. (A) pH, (B) total COD, (C) total solids (TS), (D) volatile solids (VS). Operational stages correspond to the periods (■) Start-up, (■) Transition, and (■) Steady. Data within dotted red lines has been excluded from the data analysis due to operational complications

98 A.1.3 COD balance

Figure A-III. Evolution of the COD mas balance profiles of the 4 digesters (♦) SL, (♦) SS, (♦) PL, (♦)BG. Values close to 0 indicate that the theoretical COD removal calculated from the methane production equals to the experimental COD removal. Operational stages correspond to the periods (■) Start-up, (■) Transition, and (■) Steady. Data within dotted red lines has been excluded from the data analysis due to operational complications.

99 A.2 Relative metabolic rates A C

B D

100 A.3 Microbial communities

A.3.1 NMDS and compositional dissimilarity based on Bray-Curtis distance A B

Figure A-V. (A) NMDS of the microbial community profiles at the OTU level (Bray-Curtis dissimilarity, stress = 0.171) for the four digesters over 10 sampling events (day 0 to 295). Each sample is represented by a single circle, coloured according to digester designations in previous plots. Circle size increases with time. (B) Compositional dissimilarity (Bray-Curtis distance) between the microbial communities in the four digesters at each time point relative to the initial community. (C) Compositional dissimilarity (Bray-Curtis distance) between the microbial communities in the four digesters at each time point relative to the previous sampled community. Operational stages correspond to the periods (■) Start-up, (■) Transition, and (■) Steady.

101 A.3.2 Relative abundance explained by unique/shared OTUs over time

Total Community Kingdom Bacteria Kingdom Archaea

Figure A-VI. % Relative Abundance (RA) explained by unique OTUs (Coloured), shared between 2 digesters (light grey), shared between 3 digesters (dark grey), and shared among the 4 digesters (black). In Bacterial kingdom brown shadowed area represents the %RA of archaeal populations. In Archaeal kingdom, pale-blue shadowed area represents the % RA belonging to bacteria. Operational stages correspond to the periods (■) Start-up, (■) Transition, and (■) Steady.

102 A.3.3 Analysis of Procrustes rotations based on individual PCA Indicidual PCA ordinations for each digester

Procrustes rotations

Figure A-VII. Individual PCA ordinations for each digester and Procrustes analyses at OTU level (Hellinger transformed rarefied data). Each coloured circle represents one digester, and circle size increases with operational time. The axes and arrows indicate the translation and rotation of the PCA plots (pairwise comparison)

103 Table A-I. Summary statistics from pairwise Procrustes analyses.

Sum of Squares Correlation Significance (P) (m12 squared) (symmetric rotation) SL-SS 0.0672 0.9658 0.001 SL-PL 0.2332 0.8756 0.002 SL-BG 0.2199 0.8832 0.001 SS-PL 0.1812 0.9049 0.001 SS-BG 0.1458 0.9242 0.001 PL-BG 0.0582 0.9705 0.001

104 A.3.4 Heatmap

Figure A-VIII. Heatmap 97% similarity. Non-filtered taxa for the four digesters at the 10 sampling points. Right column shows the lowest possible taxonomic assignment, and phylum level at left column. Darker intensity indicates higher relative abundance, grey cells indicate that this taxa was not detected in the samples at that level of resolution

105 A.3.5 Correlation map metabolic rates-taxa

Figure A-IX. Complete correlation maps taxa – metabolic activity rate. Left Pearson correlations, Right Spearman correlations.

106 A.3.6 Microbial community structure over time

Figure A-X. Evolution of -diversity indexes with time for the four digesters (♦) SL, (♦) SS, (♦) PL, (♦) BG. Day 0 represents the original inocula. (A) Estimated richness based on 16S OTUs clustered at 97% similarity. (B) Simpson index (0 represents complete dominance and 1 complete evenness). (C) Shannon-Wiener Entropy index.

107

Appendix B

Appendix B: Supplementary material for Chapter 6

108 B.1 Digesters’ performance

B.1.1 Individual VFA yields B.1.1.1 Major VFA A

Figure B-I. Individual VFA yields of the four digesters (♦) SL, (♦) SS, (♦) PL, (♦) BG at each SRT (■) 15-d and (■) 8-d SRT, (■) 4-d and (■) 2-d SRT. (A) Acetate and (B) Propionate. Data within dotted red lines has been excluded from the data analysis due to operational complications

109 B.1.1.2 Minor VFA A B

C D

Figure B-II. Individual VFA yields of the four digesters (♦) SL, (♦) SS, (♦) PL, (♦) BG at each SRT (■) 15-d and (■) 8-d SRT, (■) 4-d and (■) 2-d SRT. (A) i-Butyrate, (B) n-Butyrate, (C) i-Valerate, (D) n-Valerate, and (E) Caproate. Data within dotted red lines has been excluded from the data analysis due to operational complications

110 B.1.2 TS, VS, tCOD, sCOD, and pH A B

C D

Figure B-III. Monitoring parameters of the four digesters (♦) SL, (♦) SS, (♦) PL, (♦) BG at each SRT (■) 15-d and (■) 8-d SRT, (■) 4-d and (■) 2-d SRT. (A) TS, (B) VS, (C) tCOD, (D) sCOD, and (E) pH. Data within dotted red lines has been excluded from the data analysis due to operational complications

111 B.1.3 Summary of operational conditions and process performance

Table B-I. Operational conditions and process performance for each digester at steady state conditions at each SRT (average ± standard deviation)

Digester SL SS PL BG

Operational Conditions SRT 15 8 4 2 15 8 4 2 15 8 4 2 15 8 4 2 (d) Digester volume 4 3 3 3 4 3 3 3 4 3 3 3 4 3 3 3 (L) Influent flow rate 254 ± 12 360 ± 15 731 ± 12 1479 ± 11 261 ± 5 373 ± 11 735 ± 5 1506 ± 17 260 ± 6 372 ± 9 734 ± 4 1509 ± 16 260 ± 6 370 ± 9 728 ± 8 1498 ± 22 (mL d-1) OLR 0.96 ± 0.05 0.96 ± 0.04 0.98 ± 0.04 0.99 ± 0.01 0.99 ± 0.02 0.99 ± 0.03 1.01 ± 0.02 1.00 ± 0.01 0.99 ± 0.02 0.99 ± 0.03 1.01 ± 0.02 1.01 ± 0.01 0.98 ± 0.02 0.99 ± 0.02 1.01 ± 0.03 1.00 ± 0.01 (gCOD Lr-1 d-1) Process performance TS 6.0 ± 0.3 4.0 ± 0.3 2.8 ± 0.1 1.5 ± 0.1 6.1 ± 0.3 4.2 ± 0.1 3.0 ± 0.3 1.8 ± 0.1 6.1 ± 0.2 4.2 ± 0.1 2.7 ± 0.1 1.7 ± 0.1 6.1 ± 0.4 4.1 ± 0.2 2.7 ± 0.1 1.6 ± 0.1 (g L-1) VS 2.8 ± 0.2 2.3 ± 0.1 1.5 ± 0.1 0.47 ± 0.01 2.9 ± 0.3 2.3 ± 0.2 1.6 ± 0.1 0.81 ± 0.02 2.9 ± 0.2 2.2 ± 0.2 1.4 ± 0.1 0.69 ± 0.03 2.9 ± 0.3 2.1 ± 0.2 1.3 ± 0.1 0.64 ± 0.02 (g L-1) tCOD 3.0 ± 0.2 2.3 ± 0.1 1.6 ± 0.1 0.75 ± 0.06 2.9 ± 0.2 2.3 ± 0.1 1.9 ± 0.1 1.13 ± 0.03 3.0 ± 0.3 2.3 ± 0.1 1.4 ± 0.1 0.81 ± 0.06 3.0 ± 0.2 2.2 ± 0.1 1.4 ± 0.1 0.80 ± 0.04 (g COD L-1) sCOD 218 ± 77 198 ± 24 174 ± 28 143 ± 11 243 ± 78 168 ± 13 214 ± 24 91 ± 18 219 ± 108 237 ± 24 188 ± 11 52 ± 9 312 ± 81 162 ± 11 86 ± 19 36 ± 9 (mg COD L-1) tVFA -1 18.0 ± 4.9 18.0 ± 3.8 72.4 ± 7.3 55.1 ± 4.0 18.7 ± 5.6 11.8 ± 2.2 138 ± 12 67.2 ± 8.2 18.6 ± 3.7 13.0 ± 3.1 69.6 ± 7.1 28.8 ± 4.8 21.0 ± 6.2 10.2 ± 2.3 21.6 ± 6.1 11.7 ± 2.4 (mg CODeq L ) Acetate -1 12.4 ± 3.0 12.3 ± 1.9 34.5 ± 1.7 13.2 ± 1.4 12.1 ± 2.5 7.4 ± 1.2 72.7 ± 4.6 33.7 ± 6.0 10.9 ± 2.6 8.3 ± 1.3 20.8 ± 2.9 9.8 ± 1.4 14.6 ± 4.0 7.7 ± 1.8 11.5 ± 3.1 8.0 ± 2.8 (mg CODeq L ) Propionate -1 2.1 ± 0.7 3.5 ± 1.3 30.3 ± 3.8 33.5 ± 1.4 2.3 ± 1.0 1.8 ± 0.6 51.6 ± 6.1 25.6 ± 1.9 2.9 ± 1.0 1.3 ± 1.1 40.9 ± 8.2 14.2 ± 3.0 3.0 ± 2.0 1.0 ± 0.6 7.8 ± 3.3 2.2 ± 1.4 (mg CODeq L ) Butyrate -1 1.4 ± 0.5 1.0 ± 0.5 2.9 ± 1.2 4.2 ± 1.7 1.5 ± 0.6 0.8 ± 0.3 7.6 ± 3.2 3.5 ± 1.8 1.8 ± 0.5 0.7 ± 0.3 3.7 ± 1.3 1.5 ± 0.5 1.9± 0.7 0.7 ± 0.3 1.2 ± 0.3 0.5 ± 0.3 (mg CODeq L ) Valerate -1 1.5 ± 0.6 1.2 ± 0.9 4.4 ± 2.0 4.3 ± 1.9 1.6 ± 0.8 1.0 ± 0.4 5.4 ± 2.3 4.1 ± 1.8 2.5 ± 1.1 1.0 ± 0.6 3.2 ± 1.0 2.2 ± 0.7 2.2 ± 1.8 0.7 ± 0.2 1.1 ± 0.4 0.7 ± 0.3 (mg CODeq L ) pH 7.3 ± 0.1 6.9 ± 0.1 6.7 ± 0.1 6.5 ± 0.1 7.3 ± 0.1 6.9 ± 0.1 6.6 ± 0.1 6.5 ± 0.1 7.3 ± 0.1 6.9 ± 0.1 6.7 ± 0.1 6.5 ± 0.1 7.3 ± 0.1 6.9 ± 0.1 6.7 ± 0.1 6.5 ± 0.1 (-) Methane yield -1 277 ± 21 240 ± 15 214 ± 13 163 ± 9 280 ± 23 226 ± 16 174 ± 9 126 ± 5 278 ± 26 239 ± 15 223 ± 12 172 ± 7 281 ± 22 237 ± 18 229 ± 13 167 ± 9 (L CH4 kgCODin ) CH content 4 53 ± 2 52 ± 1 56 ± 3 60 ± 1 53 ± 2 52 ± 1 56 ± 3 62 ± 2 53 ± 1 52 ± 1 57 ± 2 62 ± 2 53 ± 1 52 ± 1 56 ± 2 61 ± 3 (%) COD removad 79 ± 3 71 ± 2 61 ± 1 62 ± 3 81 ± 3 71 ± 2 52 ± 1 43 ± 2 80 ± 3 71 ± 1 65 ± 2 60 ± 3 80 ± 2 73 ± 2 65 ± 1 60 ± 2 (%)

112 B.1.4 COD balance

Figure B-IV. Evolution of the COD mas balance profiles of the four digesters (♦) SL, (♦) SS, (♦) PL, (♦) BG at each SRT (■) 15-d and (■) 8-d SRT, (■) 4-d and (■) 2-d SRT. Values close to 0 indicate that the theoretical COD removal calculated from the methane production equals to the experimental COD removal. Data within dotted red lines has been excluded from the data analysis due to operational complications.

113 B.2 Relative metabolic rates A B

C D

E F

114 Continuation Figure SX H I

115 B.3 Microbial communities

B.3.1 PERMANOVA on PCA with SRT as a factor

Figure B-VI. Visualisation of PERMANOVA results testing differences among SRT (groups) on the microbial populations (OTUs Hellinger transformed) of the four digesters. Two possible clusters considering the centroid and dispersion were observed (■) 15-d and (■) 8-d SRT, (■) 4-d and (■) 2-d SRT.

116 B.3.2 Individual PCA A B

C D

Figure B-VII. PCA of the microbial community profiles at the OTU level (Hellinger transformed) for the four digesters individually over 16 sampling events (4 at each SRT). Each sample is represented by a single circle, coloured according to digester designations in previous plots. Circle size decreases for each SRT. OTUs are presented as black crosses and the populations contributing most to the variability between microbial communities are identified. Each sample is represented by a single circle, coloured according to digester designations in previous plots (A) SL, (B) SS, (C) PL, and (D) BG.

117 B.3.3 Bacterial and Archaeal Kingdom PCA A

Figure B-VIII. PCA of the microbial community profiles at the OTU level (Hellinger transformed) for the four digesters over 16 sampling events (4 at each SRT). Each sample is represented by a single circle, coloured according to digester designations in previous plots. Circle size decreases for each SRT. OTUs are presented as black crosses and the populations contributing most to the variability between microbial communities are identified. Correlation with performance parameters are indicated by arrows, blue arrows represent those parameters correlated with a significance lower than 0.01, and orange arrows parameters with a significance lower than 0.05. (A) Bacterial community. (B) Archaeal community.

118 B.3.4 Heatmap

Figure B-IX. Microbial composition at 97% similarity (relative abundance), observed OTUs, and diversity indexes (Shannon and Simpson index). Heatmap shows the populations in the four digesters at the 17 sampling points. The lowest possible taxonomic assignment is shown in the right column and phylum level in the left column. Microbial composition: Darker intensity indicates higher relative abundance, grey cells indicate that this taxa was not detected in the samples at the level of resolution (>0.005% of total counts at least in one sample). Observed OTUs: Darker intensity indicates higher observed OTUs per sample. Diversity indexes: Darker intensity indicates greater evenness (microbial populations more evenly distributed).

119 B.4 Biofilm pictures A

Figure B-X. Biofilm pictures taken during digesters dismantling (day 401, cycle 36) of the four digesters (A) SL, (B) SS, (C) PL, and (D) BG. Details of covering walls, paddle or aggregates that came of the paddle during the opening process.

120 B.5 First-order fit cellulose hydrolysis and CSTR A

Figure B-XI. (A) First-order CSTR model fitting methane yield as function of SRT. Marker represent the average experimental methane yield at each SRT. Solid line represents the modelled profile. Each colour represents one digester (♦) SL, (♦) SS, (♦) PL and (♦) BG. (B) Comparison

of first order kinetic constant (kh) obtained from the activity assays (green) vs the kinetic constant derived from the first-order CSTR model (black). Solid line represents the value of the kinetic constant, shadowed box represents the 95% confidence interval of the value

121

Appendix C

Appendix C: Supplementary material for Chapter 7

122 C.1 Cumulative methane production curves and confidence regions

20 ºC Semi-Aerobic (1) 20 ºC Semi-Aerobic (1)

20 ºC Semi-Aerobic (2) 20 ºC Semi-Aerobic (2)

123 Figure S-XIX Continuation

20 ºC Anaerobic 20 ºC Anaerobic

37 ºC Anaerobic 37 ºC Anaerobic

124 Figure S-XIX Continuation

55 ºC Anaerobic 55 ºC Anaerobic

70 ºC Anaerobic 70 ºC Anaerobic

Figure C-I. (Left) Cumulative specific methane production curves after pre-fermentation at different temperatures, treatment time, and control. (Right) Confidence regions from BMP test modelling after pre-fermentation and control. Each ellipse bounds the confidence region (95%) of each trial for the anaerobic biodegradability (f, x-axis) and the hydrolysis rate (khyd, y-axis)

125 C.2 COD balance fermentation unit followed by anaerobic digestion

Table C-I. COD balance for the combined fermentation and anaerobic digestion process COD COD COD tCODin COD loss pCODout SCODout fi recovered lost removed (g L-1) (g L-1) (g L-1) (g L-1) (-) (%) (%) (%) Raw PS (Batch 1) 57.9 0.0 55.9 2.1 0.68 69.1 0 69.1 20 ºC Semi-aerobic 12 h 57.9 0.6 54.6 2.7 0.66 66.8 1.1 67.9 24 h 57.9 1.1 53.8 3.1 0.69 69.4 1.8 71.2 48 h 57.9 3.1 50.9 3.9 0.73 70.9 5.3 76.3 72 h 57.9 3.0 50.9 4.1 0.78 75.5 5.1 80.7 20 ºC Anaerobic 12 h 57.9 0.0 56.9 2.7 0.67 70.6 0.0 70.6 24 h 57.9 2.7 52.2 3.1 0.71 69.3 4.6 73.9 48 h 57.9 4.6 48.9 4.4 0.68 65.0 7.9 73.0 72 h 57.9 5.4 47.0 5.5 0.71 67.1 9.4 76.5 37 ºC Anaerobic 12 h 57.9 0.0 54.1 4.1 0.66 68.6 0.0 68.6 24 h 57.9 1.7 50.6 5.6 0.69 69.9 3.0 72.9 48 h 57.9 3.1 46.7 8.1 0.70 70.5 5.3 75.8 72 h 57.9 8.2 40.6 9.1 0.71 65.5 14.2 79.7 55 ºC Anaerobic 12 h 57.9 0.0 52.2 6.7 0.65 70.2 0.0 70.2 24 h 57.9 0.0 51.3 7.1 0.66 70.8 0.0 70.8 48 h 57.9 0.0 49.4 8.5 0.66 71.0 0.0 71.0 72 h 57.9 1.2 48.2 8.5 0.65 68.8 2.0 70.9 70 ºC Anaerobic 12 h 57.9 0.0 50.9 7.3 0.66 70.6 0.0 70.6 24 h 57.9 0.8 49.2 7.9 0.67 70.6 1.3 72.0 48 h 57.9 0.0 49.1 9.4 0.65 71.3 0.0 71.3 72 h 57.9 0.3 48.4 9.2 0.66 71.0 0.5 71.6

Raw PS (Batch 2) 72.1 0.0 69.6 2.5 0.54 55.6 0.0 55.6 20 ºC Semi-aerobic 24 h 72.1 1.8 66.6 3.7 0.55 55.9 2.6 58.5 48 h 72.1 2.7 65.5 3.9 0.58 58.1 3.7 61.9 72 h 72.1 0.4 67.5 4.3 0.57 59.3 0.5 59.8 96 h 72.1 0.5 66.8 4.8 0.58 60.3 0.8 61.1

126

127