1

Effects of carbon source and particle size on nitrogen cycling in aggregated “Bio-Floc” microbial communities Chelsea Westra Hampshire College 893 West Street, Amherst MA 01002 Advisor: Dr. Joe Vallino Marine Biological Laboratory 7 MBL St Woods Hole, MA 02543 Semester in Environmental Science Independent Project, 2013

2

Abstract The addition of a carbon source in aquaculture systems can stimulate microbial flocs which take up fish waste and subsequently convert it into a protein source for fish. Microbes utilize nitrogen in different ways depending on the carbon source used; past studies have reported higher rates of N immobilization in flocs treated with glycerol rather than glucose. The proposed hypothesis aims to investigate the effects C sources may have on respiration rates, which would affect N conversion pathways. In this experiment I observed the effects of C source and particle size on respiration rates and nitrogen cycling. I measured oxygen uptake rates (OUR) and N forms NH4, NO3, TDN, and PON to complete a mass balance and find N2 by difference. I also ran a protein assay to confirm higher protein concentration in the glycerol treatment. While results replicated past studies and the glycerol treatment increased protein content, differences between respiration rates and denitrification rates were difficult to determine between treatments during the given time period. Results did show significant protein increase and NH4 removal by glycerol and suggests further research on the mechanism behind this. Microscopy images show stark differences in floc composition between treatments, with hyphal growth dominating glycerol treatments, which may be key to understanding the role of carbon in nitrogen uptake.

Keywords: Biofloc technology, organic carbon source, bioreactor, nitrogen removal, denitrification process, protein concentration, bacterial aggregates

Introduction Flocculated microbial communities—aggregated microbial communities composed of heterotrophic bacteria, dead cells and polymers— have long been implemented in biological wastewater treatment, and this technique is now being used to treat nutrient buildup in aquaculture systems. Biofloc Technology (BFT) aquaculture is the cohabitation of microbial communities and fish that aims to solve two major issues in fish production: wastewater treatment and protein addition. Conventional aquaculture requires constant replacement of freshwater to prevent toxic waste buildup, utilizing a scarce resource and producing polluted effluent (Schryver et al 2008). The second major input into aquaculture systems is the need for high-quality protein, and fisheries are generally dependent on external fishmeal to provide necessary nutrients (Tacon and Metian 2008). As the aquaculture industry continues to expand, pressure on fishmeal production will increase at an unsustainable rate (Hardy 1996; Carter & Hauler 2000).

3 BFT aquaculture aims to remediate these issues with the presence of floc to convert fish waste into protein, thereby removing dissolved nutrients from the system and converting waste into protein available for fish consumption. In these systems, an external carbon source is required and is processed by bacteria in the following manner: Organic C  CO2 + energy + C assimilated in microbial cells Microbial conversion efficiency of C ranges from 40-60% and this uptake requires nitrogen for protein (Paul and Van Veen 1978). The carbon source serves as a carbohydrate that encourages N immobilization, removing toxic inorganic forms and creating protein. Optimal carbon to nitrogen ratios in these systems are >10, with nitrogen provided by fish waste and external carbon inputs to maintain flocs and N uptake (Avnimelech 2009). Efficient engineering that creates an appropriate balance between C and N can produce the most efficient system to purify water and create protein. It is therefore important to understand nutrient cycling dynamics within these flocs to optimize their performace in aquaculture systems (Schryver et al 2008). Crab et al (2010) found that there was a significant increase in protein content of floc fed glycerol as opposed to glucose, despite their similar energy content with aerobic metabolism. This study showed that different carbon sources alter floc bacterial structure, which in turn affects the N processes occurring within the flocs (Schryver et al 2008). It is important to understand how and why nitrogen immobilization varies with carbon source in order to optimize N immobilization in BFT aquaculture (Crab et al 2012); therefore, in this experiment I aimed to understand nutrient cycling dynamics within floc. My hypothesis observes the effects of C sources on respiration rates and how this affects N conversion pathways. I track respiration because increased respiration in

4 the system could promote the presence of anoxic pockets within floc. These anoxic micro-niches within floc could support denitrifying bacteria that produce N2 gas, an anaerobic process that removes N from the system. This export of N from the system into the atmosphere leads to lower immobilization rates and decreased protein within flocs. Particle size could also affect denitrification, as larger particles are more likely to host anoxic pockets and promote N2 production. My project observes the relationship between respiration rates and N conversion. My hypothesis centers around the idea that glucose addition leads to higher respiration rates than glycerol, increasing denitrification rates and N export from the system. I also manipulate particle size to observe N cycling and respiration rates with varied particle size.

Methods I set up eight two-liter graduated cylinders each with a working volume of 1.5 liters. Four treatments were run in duplicate with two carbon sources, glucose and glycerol, and small and large particle size. Weighted air stones placed at the bottom of each graduated cylinder ensured constant suspension of floc particles and aerobic conditions. The bioreactors were kept in a dark growth chamber to discourage growth of photosynthetic organisms and temperature was kept at 30 C. The systems were initially inoculated with 100 mL of biofloc from a BFT tilapia system at the Woods Hole Oceanographic Institution. Initial N concentration of each system was 1.8 mM NH4. The systems were run as pulse chemostat, which entailed 10% volume exchange daily. Added media was composed of 18 mM (NH4)2SO4 and corresponding C

5 concentration to attain a C:N ratio of 11. KH2PO4 and 60 mL of water from John’s Pond in Falmouth, MA were also added to provide trace minerals and avoid nutrient limitations. I monitored pH three times a day, adding 1 N HCl and sodium bicarbonate as needed to maintain pH levels around 8.5. I used an immersion blender to agitate the unflocculated systems twice daily to discourage flocculation and decrease particle size. Removed water was filtered through ashed 47 GF/F filters. Filters were dried at 50 C for 24 hours and stored in a desiccator. Ammonium samples were preserved with 5N HCl and samples were frozen for nitrate (NO3) and Total Dissolved Nitrogen (TDN) analyses. Samples were periodically analyzed under a Zeiss microscope to determine particle size distribution. I determined oxygen uptake rates (OUR) using a WTW Dissolved Oxygen probe throughout the experiment. Rates were determined by monitoring O2 decrease in the absence of oxygenation, which was collected 5 hours after media input. I used uptake rates to determine steady state and nutrient analyses were focused on 5 days, from November 27 to December 2. GF/F filters were run on the CHN elemental analyzer (Perkin-Elmer) to determine particulate organic N and C. NH4 analysis followed methods modified from Strickland and Parsons and samples were run on the Shimadzu UV-1601 spectrophotometer at 200:1 dilution (Strickland and Parsons 1972). NO3 was run on the Lachat via automated colorimetric flow injection analysis following the QuikChem Method at 20:1 dilution (Diamond 2008). These two analyses comprised dissolved inorganic N measurements.

6 To determine TDN, I added potassium persulfate to samples diluted 50:1, which were then autoclaved for 1.5 hours. Once nitrogen was fully oxidized into nitrate, samples were run on the Lachat using nitrate methods listed above. I determined protein concentrations by sonicating samples for three minutes to encourage cell lysis for a protein assay (Sigma Aldrich 51254). Samples were dyed with Coomassie Brilliant Blue G (CBB) and run in a 96 well microplate on the Spectramax Plus 384 with bovine serum albumin (BSA) as the standard. Respiration rates were determined from rate of oxygen uptake by finding the slope of initial curve, as described by Hagman and Jansen (2007) (Fig. 1). N mass balance was completed using PON, DIN and TDN measurements. N2 production was found via the difference. Rates of ammonium uptake, nitrification, denitrification and N immobilization were extrapolated from this data given Equations 1 & 2.

Results Both carbon sources were able to uptake the majority of ammonium. Figure 2 shows projected ammonium concentrations given a sterile system scenario, and treatments had converted >11 mM NH4 at initial analysis. Flocculated glycerol had the highest removal rate, with an average of 95.4  0.4% NH4 removal over the 5 day period. Flocculated glucose removed the least NH4 with an average of 88.3  1.9% NH4 removal. Ammonium concentrations were therefore lowest in the flocculated glycerol treatment throughout the 5 day sampling period, with an average ammonium concentration of 0.66  0.05 mM NH4 (Fig. 3). Flocculated glucose systems had the highest initial ammonium levels and displayed ammonium buildup, increasing from 1.37 to 2.03 mM over 5 days.

7 Protein concentrations are displayed in Figure 4. Unflocculated glycerol treatment had the highest protein content, with 135.5  15.5 µg/ml. Unflocculated treatments in both C sources displayed higher protein content than their flocculated counterparts—33.8  8 compared with 88.3  23.5 µg/ml for glucose and 65.9  9.7 to 135.5  15.5 µg/ml for glycerol. Respiration rates throughout the experiment and within the steady state period are displayed in Figures 5a and 5b. Although initial glucose respiration rates were significantly higher than glycerol—177.2  17 mmol O2 liter-1 d-1 in glucose-F compared with 21.5  7.5 mmol O2 liter-1 d-1 on day 1—respiration rates varied greatly as microcosms established. Respiration rates ranged from 12.1  1 to 77.5  56.3 mmol O2 liter -1 d-1 during the analysis period—fluctuations between duplicates and microcosms did not allow for any detectable trends or distinctions between treatments. Particulate organic N concentration was significantly higher in flocculated glycerol than glucose, with an average 241.5  15.4 compared with 162.4  38.8 µg/ml (Fig. 6). Unflocculated glucose treatments had the lowest concentrations with an average 59.3  21.1 µg/ml. Molar C:N ratios were lowest in glycerol treatments and decreased throughout the analysis in all treatments (Fig. 7). Nitrate levels were nearly undetectable during the sampling period, although were initially highest in glycerol treatments before decreasing over time (Fig. 8). Total dissolved N concentrations reflected NH4 data, suggesting majority of dissolved N was in the form of NH4. N mass balance results derived from rate calculations are shown in Figure 9. Negative NH4 rates denote ammonium uptake, while positive PON and N2 rates convey

8 N immobilization and denitrification rates. All treatments show consistent ammonium uptake and occurrence of denitrification. Unflocculated glycerol treatments had highest rates of N immobilization with 1.57 mmol PON liter-1 d-1. Figure 10 shows the relationship between denitrification rate and protein content. Particle size distribution is compared between treatments in Figure 11. Particles sized between 101-300 µm were most abundant in flocculated glucose treatments, whereas glycerol treatments had more particles in the 401-600 µm range (Fig. 11a). Unflocculated treatments had similar distributions between C treatments and were clustered between 0-200 µm (Fig. 11b). Microscopy images are displayed in Figures 12 and 13 to show differences in floc composition between carbon sources.

Discussion All systems proved to be efficient at NH4 removal. Decreasing C:N ratios among all treatments shows increasing N immobilization throughout the sampling period. Results from the protein assay confirm that findings in Crab et al (2010) were replicable and glycerol had higher protein concentrations. Particulate N data also supports this, as glycerol treatments had higher levels of N accumulation. Glycerol also displayed higher rates of NH4 uptake than glucose. This suggests that glycerol addition promotes increased microbial N uptake and immobilization. While glucose initially had higher respiration rates, it is difficult to make any conclusive remarks given the variability between duplicates and treatments. The systems may have still been reaching steady state, and running the experiment for longer might produce more clear trends. Although it is not possible to determine a relationship between respiration and denitrification rates, results from mass balance calculations suggest

9 denitrification is likely occurring in most treatments. This supports the idea that anaerobic microniches are present within the floc. However, variations throughout treatments suggest that the C source did not have a large effect on denitrification rates. Although total Particulate N was higher in flocculated treatments, protein concentration and immobilization rates were higher in unflocculated treatments for both C sources. Despite the most total NH4 removal in flocculated glycerol, unflocculated glycerol had the highest N immobilization rate, highest protein concentration, and lowest denitrification rates. This could support the current hypothesis; smaller particle size might discourage development of anaerobic pockets, thereby decreasing denitrification rates. Most related research has been done on sequencing batch reactors (SBRs) in wastewater treatment plants that support aerobic/anaerobic stages, so future research could focus on micro-habitats within floc that are influenced by microbial respiration rather than external oxygen supply. The most obvious difference between carbon sources was observational data taken from microscopy analysis. Glycerol treatments showed hyphal structures dominating flocs that were mostly absent in glucose. Crab et al (2010) suggest that C sources may affect microbial composition which affects N immobilization; glucose may promote floc that expend more energy in producing exopolysaccharides while glycerol encourages bacteria that immobilize more N as protein. Future research should concentrate on the effects of C sources on floc community composition. Understanding how small differences among the carbohydrate source could greatly affect community composition would give insight into nutrient cycling dynamics

10 within microbial communities. This would help inform decisions to optimize the use of flocculated microbes for biological remediation of wastewater.

Acknowledgements Firstly to Joe Vallino for all the guidance and troubleshooting along the way. The project wouldn’t have materialized without the knowledge and inoculant from Bill Mebane, who also helped keep the big picture in mind. Thanks to Sarah Nalven and Rich McHorney because I don’t think I would have finished without their support and reassurance throughout. Alice Carter and Fiona Jevon for engaging in biofloc discussions towards the end and helping me pull it all together. Thanks to Ken Foreman for directing such a great program and to everyone involved in the SES program. References: Avnimelech, Y. 2009. Biofloc Technology: a practical guidebook. World Aquaculture Society. pp 1-42 Carter, C G & Hauler R C. 2000. Fish meal replacement by plant meals in extruded feeds for Atlantic salmon, Salmo salar L. Aquaculture 185, 299-311. Crab R, Defoirdt T, Bossier P, Verstraete, W. 2012. Biofloc technology in aquaculture: beneficial effects and future challenges. Aquaculture. 351-356. Crab R, Chielens B, Wille M, Bossier P, Verstraete W. 2010. The effect of different carbon sources on the nutritional value of bioflocs, a feed for Macrobrachium rosenbergii postlarvae. Aquaculture Research. 41, 559-567 Diamond D. 2008. Determination of nitrate and/or nitrite in brackish or seawater by flow injection analysis colorimetry. Lachat Instruments. Loveland, CO. Hardy, R. W. 1996. Alternate protein sources for salmon and trout diets. Animal Feed Science and Technology.59, 71-80. Hagman, M & Jansen J. 2007. Oxygen uptake rate measurements for application at wastewater treatment plants. Water and Environmental Engineering. 63, 131-138. Paul, E.A., van Veen, J.A., 1978. The use of tracer to determine the dynamic nature of organic matter. Proceedings of the 11th International Congress of Soil Science, Edmonton, Canada. 3, 61–102. Schryver, P D, Crab R, Defoirdt T, Boon N, Verstraete W. 2008. The basics of bio-flocs technology: the added value for aquaculture. Aquaculture 277, 125-137. Strickland, J D H & Parsons T R. 1972. A Practical Handbook of Seawater Analysis. Fisheries Research Board of Canada. 2nd ed. Ontario, Canada. Tacon A G, Metian M. 2008. Global overview on the use of fish meal and fish oil in industrially compounded aquafeeds: trends and future prospects. Aquaculture. 285, 146-158.

11 Equations:



Cti1 Cti f  (Cf  Cti ) ti1  ti V

(1)

RN 

(2)

RN2 

(RNH4  RNO3  RPON  RDON ) 2

 Figures: Fig. 1. Oxygen data from Glyc-UF2 on Dec 2 exemplifies how oxygen uptake rates (OUR) were determined. Dissolved oxygen concentration generally decreased in a linear fashion so that the slope constituted OUR. Fig. 2. Projected NH4 concentration in sterile scenario where no ammonium is taken up. Fig. 3. Actual NH4 concentrations across treatments throughout the steady state period. Fig. 4. Results from the protein assay showing higher concentrations in unflocculated treatments. Fig. 5. Respiration rates throughout the entire experiment (5a) and during steady state analysis (5b). Fig. 6. Particulate N concentrations across treatments. Fig. 7. Molar C:N ratio throughout steady state which shows increasing N content over time. Fig. 8. Nitrate levels were very low throughout steady state in all treatments (note concentrations are in µM). Fig. 9. Visual representation of the N mass balance, showing steady ammonium uptake rates and denitrification likely occurring in all treatments. Fig, 10. The relationship between denitrification rates and protein content is unclear, but glycerol-UF displayed the highest protein concentration and lowest denitrification rate. Fig 11. Particle size distributions across treatments. 11a shows the differences in size between flocculated glucose and glycerol, while 11b displays the particle size uniformity in unflocculated systems. Fig. 12. Microscopy image of flocs with glucose addition. Fig 13. Flocs in glycerol treatments were dominated by hyphal structures that were mostly absent with glucose addition.

Dissolved Oxygen (mg/l)

12 6.7 6.6 6.5 6.4 6.3 6.2 6.1 6 5.9 5.8

y = -0.3144x + 6.6458 R² = 0.9993

0

0.5

1

1.5 Time (min)

2

2.5

3

Fig. 1 Oxygen data from Glyc-UF2 on Dec 2 exemplifies how oxygen uptake rates (OUR) were determined. Dissolved oxygen concentration generally decreased in a linear fashion so that the slope constituted OUR.

13

NH4 concentration (mM)

18 16 14 12 10 8 6 4

2 0 13-Nov

18-Nov

23-Nov

28-Nov

3-Dec

Fig. 2 Projected NH4 concentration in sterile scenario where no ammonium is taken up.

14 2.50

NH4 Concentration (mM)

2.00

1.50

Glu-F Glu-UF Glyc-F

1.00

Glyc-UF

0.50

0.00 27-Nov

28-Nov

29-Nov

30-Nov

1-Dec

Fig. 3 Actual NH4 concentrations across treatments throughout the steady state period.

15

Protein concentration (µg/ml)

160 140 120

100 Flocculated

80

Unflocculated

60 40 20 0

Glucose

Glycerol

Fig. 4 Results from the protein assay show higher concentrations in unflocculated treatments.

16

Respiration Rate (mmol O2 liter-1 day-1)

200 180 160 140 120

Glucose-F

100

Glucose-UF

80

Glycerol-F

60

Glycerol-UF

40 20 0 13-Nov

18-Nov

23-Nov

28-Nov

3-Dec

Respiration Rate (mmol O2 liter-1 day1)

160 140 120 100

Glucose-F Glucose-UF

80

Glycerol-F

60

Glycerol-UF

40 20

0

26-Nov

28-Nov

30-Nov

2-Dec

Fig. 5 Respiration rates throughout the entire experiment (5a) and during steady state analysis (5b).

17

300

PON (µg/ml)

250 200

Glucose-F Glucose-UF

150

Glycerol-F 100

Glycerol-UF

50 0

26-Nov

27-Nov

28-Nov

29-Nov

Fig. 6 Particulate N concentrations across treatments.

30-Nov

1-Dec

18

10 9

Molar C:N ratio

8

7 6

Glucose-F

5

Glucose-UF

4

Glycerol-F Glycerol-UF

3 2 1

0 26-Nov

27-Nov

28-Nov

29-Nov

30-Nov

1-Dec

Fig. 7 Molar C:N ratio throughout steady state which shows increasing N content over time.

19 1.6

NO3 concentration (µM)

1.4 1.2 1

Glucose-F Glucose-UF

0.8

Glycerol-F

0.6

Glycerol-UF

0.4 0.2

0 26-Nov

27-Nov

28-Nov

29-Nov

30-Nov

1-Dec

Fig. 8 Nitrate levels were very low throughout steady state in all treatments (note concentrations are in µM).

20

2.5 2.0

Δ N (mmol N liter-1 day-1)

1.5 1.0 Δ NH4 Δ NO3

0.5

Δ PON (Immobilization) 0.0 Glucose-F

Glucose-UF

Glycerol-F

Glycerol-UF

Δ N2 (Denitrification) Δ DON

-0.5 -1.0 -1.5 -2.0

Fig. 9 Visual representation of the N mass balance, showing steady ammonium uptake rates and denitrification likely occurring in all treatments.

Denitrification rate (mmol N2 l-1 d-1)

21

0.25 0.2 0.15

0.1 0.05 0 0

50 100 Protein concentration (µg/ml)

150

Fig. 10 The relationship between denitrification rates and protein content is unclear, but glycerolUF displayed the highest protein concentration and lowest denitrification rate.

Frequency

22

20 18 16 14 12 10 8 6 4 2 0

Glucose-F Glycerol-UF

Particle Size (µm) 40

35 Frequency

30 25

20 15

Glucose-UF

10

Glycerol-UF

5 0

Particle Size (µm)

Fig. 11 Particle size distributions across treatments. 11a shows the differences in size between flocculated glucose and glycerol, while 11b displays the particle size uniformity in unflocculated systems.

23

Fig. 12 Microscopy image of flocs with glucose addition.

24

Fig. 13 Flocs in glycerol treatments were dominated by hyphal structures that were mostly absent with glucose addition.