Materials and methods Directory UMM :Data Elmu:jurnal:A:Aquacultural Engineering:Vol22.Issue3.Jun2000:

provide insight to how biofilm structure develops and adapts to certain environ- mental conditions within fluidized sand filters.

2. Materials and methods

2 . 1 . Reactor set-up Twelve experimental fluidized sand filters were constructed using clear acrylic cylinders each 2.5 m tall, 16 cm diameter. These units were operated in parallel with an established fluidized sand filter 2.5 m tall, 1.5 m diameter used as the biological filter system to culture rainbow trout in a cold water 14 – 17°C system. A mechanical screen filter 110 mm removed solids before wastewater treatment by biofilters. Further solids removal occurred by a second mechanical screen 80 mm following treatment by the large biofilter. Low-head oxygen units were used to maintain the culture tanks above 10 ppm at all times see Table 1 for all influent conditions. Six of the previously mentioned twelve reactor vessels biofilters were used for this experiment. Water was introduced at the bottom of the reactor vessels by a 2.54 cm SCH-40 PVC pipe located down the center of the column. Wastewater was discharged in a downward direction opposite to the direction of outflow from the vessel approximately 8 cm above a concrete cup formed into the reactor vessel cylinder that was 10.5 cm inside diameter and had an 8 cm inside cavity depth. Each experimental filter used FilterSil sand Unimin, New Canaan, CT with effective diameter sand sizes D 10 of 0.23 or 0.60 mm. The effective diameter is defined as the opening size which will pass only the smallest 10 by weight of the total material Summerfelt and Cleasby, 1996. The uniformity coefficient for sands with D 10 0.23 and 0.60 mm were 1.8 and 1.4, respectively. Each filter containing one Table 1 The clean static sand depth, fluidized bed depth, superficial velocity, ammonia concentration, nitrite concentration, dissolved oxygen concentration, and TVS static bed concentration for reactors of each sand size 9 SE, from Tsukuda et al. 1997 a Parameters Large biofilter Experimental biofilters 0.23 Effective sand diameter mm 0.23 0.60 NA Clean static sand depth cm 14.6 9 1.9 59.0 9 0.7 90.3 9 1.6 174.7 9 8.3 NA Mean fluidized bed depth cm 2.94 9 0.05 0.82 9 0.01 0.71 SE N Superficial velocity cms A 0.49 9 0.03 0.06 9 0.01 0.11 9 0.04 Effluent ammonia concentration mgl 0.085 9 0.008 0.137 9 0.012 0.024 9 0.013 Effluent nitrite concentration mgl 7.0 9 2.0 10.28 9 0.10 6.18 9 0.27 Effluent dissolved oxygen concentration mgl 35 521 9 0.06 NA 1895 9 0.02 Total volatile solids mgl static bed height a Influent conditions all reactors mgl: 0.55 9 0.03 ammonia; 0.063 9 0.005 nitrite; 10.76 9 0.10 oxygen. of the two sand sizes was replicated three times, totaling six experimental columns. The large 1.5 m diameter biofilter consisted of 0.23 mm D 10 sand. 2 . 2 . Reactor operation and performance Clean static sand depth, fluidized bed height, flow rates, and water quality measurements were reported during weeks 35 – 42 of reactor operation and are listed in Table 1. Further information on biofilter set-up, operation, and perfor- mance can be found in Tsukuda et al. 1997. Note that the intent of this paper is to focus on biofilm characteristics and not performance. 2 . 3 . Biofilm sampling, preparation, and imaging A one-time collection of sand samples was taken during week 35 from each of the 0.23 and 0.60 mm sand reactor vessels to compare biofilm from small and large sand reactors. Samples were taken 20.5 cm low and 70.5 cm high from the bottom of the reactor to compare biofilm from low and high locations in a biofilter. A tube attached to a syringe was lowered to each sampling location to collect particles. Samples were placed in 15 ml glass vials and immediately preserved in 3.7 methanol free ultra-pure formaldehyde CH 3 OH; 16; Polyscience, Warring- ton, PA with 50 mM HEPES buffer N[2-Hydroxyethyl]piperazine-N-[2-ethane- sulfonic acid]; 99.5; Sigma, St. Louis, MO. Pure formaldehyde was chosen since it does not cause osmotically-induced swelling and lysis and the fixative penetrates specimens rapidly Hayat, 1981. Several sand particles were placed on a sterile slide and rinsed with 0.1 ml of 50 mM HEPES buffer to remove excess fixatives. Samples were stained with sterile 4 m M acridine orange base C 17 H 19 N 3 ; 95; Sigma, St. Louis, MO for 5 min. Acridine orange is a positively charged fluorescent dye that binds well to DNA and RNA. Due to its low binding specificity, acridine orange is also capable of binding to dense extracellular polymers. Samples were rinsed with a sterile 0.1 M NaCl and 50 mM HEPES buffer wash to remove excess stains. Each biofilm particle was placed in 1 mm deep, 20 mm diameter imaging chamber Grace Biolabs, Sunriver, OR. Anti-fading reagent DABCO was used to reduce photobleaching of acridine orange during the laser scanning confocal microscopy process. Ten microliters of 0.1 M DABCO 1,4-Diazabicyclo [2,2,2] octane; 99.9; SIGMA, St. Louis, MO in 50 mM HEPES buffer was applied to the particle being scanned and glycerin CH 2 OHCHOHCH 2 OH; 99.7; Fisher Chemical, Springfield, NJ was added as the mounting media. Minimal photo- bleaching was observed during the course of imaging. Ten sand particles from each sampling location 2 sand sizes × 3 replicate reactors × 2 depths × 10 particles = 120 sand particles were imaged using a MRC- 600 Lasersharp fluorescence scanning confocal microscope BioRad, Cambridge, MA mounted on a Zeiss Axiovert 10 microscope Zeiss, Thornwood, NY and equipped with a krypton-argon ion laser Ion Laser Technology, Salt Lake city, UT. Acridine orange was excited at 488 and 588 nm and a FITC barrier filter was used. Laser scanning confocal microscopy was capable of producing crisp 2D images of thick specimens at various depths by rejecting out of focus information. A computer displayed the image by mapping measured fluorescence intensity to pixel brightness. Control of imaging depth was achieved by a motor-driven focusing system. A 16 × , 0.50-numerical aperture oil immersion plan neofluor objective lens with a working distance of 0.22 mm provided a pixel resolution of 0.972 × 0.972 m m 2 . Each 2D image contained 768 × 512 pixels. A series of 2D images ranging from 25 to 129 slices, depending on film thickness and particle size at incremental depths of 0.98 mm were acquired for each biofilm coated particle. Therefore, stacked 2D images rendered a 3D image of biofilm with a voxel analogous to pixel except in 3D resolution of 0.972 × 0.972 × 0.98 mm 3 . Given the thickness of biofilm covered sand particle exceeded the 0.22 mm working distance of the objective lens, the microscope imaged the upper portion of biofilm coated sand particles. Images were continually taken at deeper depths of one biofilm coated particle until the user could distinctly recognize a circular black void sand particle surrounded by fluorescing biofilm. No signals were detected from the sand particle since fluorescence of sand was minimal and light did not pass through the sand particle. 2 . 4 . Image processing and analysis All images were processed and analyzed on a SGI ONYX Workstation Santa Clara, CA. Biofilm surface area and volume were measured for a defined sand surface and reported as biofilm volume per unit sand area and biofilm surface area per unit sand area to standardize measurements. The sand surface area used to standardize the biofilm characterizations on a per unit area basis was determined in the last 2D slice of a 3D image where the sand particle created a distinct black void. A circle to determine area was manually drawn within the perimeter of the sand particle in the last 2D image. This circular area was the projected area of the sand surface of the biofilm analyzed. The calculated sand surface area neglected surface irregularities of the sand particle. The sand area was calculated by the sum of the number of pixels in the circular area multiplied by 0.945 mm 2 area of a pixel. Voxels directly above and within the projected area retained their original grayscale intensities whereas other voxels were set to zero to eliminate voxels representing biofilm not attached to the defined sand surface. A program was written using IDL software Research Systems, Boulder, CO to determine which voxels in a 3D image represented dense slime and bacterial clusters in biofilms. The program was based on the intermean thresholding algorithm Ridler and Calvard, 1978. Biofilm volume was estimated by counting the number of voxels equal or greater than the threshold intensity multiplied by 0.926 mm 3 volume of 1 voxel. Using VoxelMath Vital Images, Fairfield, IO, the surface of the biofilm was identified by detecting and marking non-zero voxels whose six faces, twelve edges, or eight corners were touching a voxel of zero intensity. AVS software Advanced Visualiza- tion Systems, Waltham, MA generated an isosurface connecting the surface voxels. The generated isosurface had an associated list of coordinates of vertices of a polyhedron with triangular facets which are determined by the tessellation of values of the surrounding 26 voxels Montemagno and Gray, 1995. Summation of these triangular areas was a measure of biofilm surface area. 2 . 5 . Statistical analysis Analysis of variances ANOVA was used to detect differences in biofilm structure due to sand sizes 0.23 versus 0.60 mm and physical location within the reactor vessel high versus low. The biofilm characteristics of interest were volume per unit sand area, surface area per unit sand area, and surface area per biofilm volume. Analysis of variances were performed on all three parameters to detect significance due to reactor variability and depth for a particular sand size. Data from reactors of the same sand size was grouped and additional ANOVA tests were performed to detect significance of sand size on measured parameters. All tests of significance were performed at an alpha level of 0.05. Stabilized variance is a necessary requirement for ANOVA tests. However, residual versus fitted plots of biofilm volume per unit sand area and biofilm surface area per unit sand area revealed increasing variance with increasing measurement values. Similar patterns were observed by Gjaltema et al. 1997. Snedecor and Cochran 1989 suggest using a natural log transformation as an appropriate method to stabilize variance in these cases for ANOVA tests. Residual versus fitted plots and linear normality plots of transformed data confirmed stabilized variance. No transformation was necessary with biofilm surface area per biofilm volume since raw data was normally distributed with a stabilized variance.

3. Results