2.3 Image processing and data analysis

FISH images of cross-sections were obtained using an Olympus IX-81 inverted microscope equipped with a long working distance (WD 2.7-4.0) 40x objective lens (NA-0.6) and a DP73 camera (Olympus, Japan). Image processing was carried out on the Fiji software (v 2.10.0) . Rolling ball background subtraction (30 pixels) and particle size (>2x2 pixels) and circularity (>0.1) filters were used to correct for uneven background and filter out false positives from the textures left on the slices by the microtome slicing. The size and count of fluorescent colonies from each section were recorded. Colony counts were normalized to the volume of detection, as calculated by the area of each image (1.47 mm²) and the thickness of the cross-section (10 μm). Box plots were constructed in R with a median represented by a centerline, a 25th to 75th interpercentile range represented by a shaded box, and smallest/largest values within the 1.5 times interquartile range represented by upper/lower error bars. Statistical analyses were performed in R (v 4.2.2). Non-parametric tests were used to compare particle sizes. The Kruskal-Wallis one-way ANOVA was conducted on triplicates at each depth, whereas the Wilcoxon signed-rank test was conducted between paired samples before and after each incubation. Student t-tests were performed on qPCR results to compare gene copy numbers before and after incubation.

2.4 DNA extraction and qPCR

To extract DNA, an entire alginate bead was dissolved in 1 mL of sodium citrate (55 mM) and EDTA (30 mM) solution. The dissolved bead solution was centrifuged at 4 °C to obtain cell pellets, which were washed twice and resuspended in buffer solution (10 mM Tris, 0.1 mM EDTA) for storage at –20 °C. The PowerSoil DNA extraction kit (Qiagen) was used to extract DNA from the pellet in the buffer. qPCR was used to quantify total bacteria, AOB, anammox bacteria, and NOB with 341F and 805R primers, amoA_1F and amoA_2R primers, hzsA_1597F and hzsA_1857R primers, and nxrB169f and mxrB638r primers, respectively (Table S1). The qPCR was performed on a StepOnePlus Real-Time PCR System (Thermo Fisher) with the standard curve method as described previously . The amplification efficiencies for all primer sets were >90%, except for the 16S rRNA gene (>85%).

Method validation with encapsulated Nitrosomonas

The background noise of FISH probes and autofluorescence of alginate embedded in GMA were first examined using abiotic alginate beads to which no microbial inoculum was added. No fluorescence was detected for all the probes used, except for CY3. The texture of GMA-embedded alginate sections emitted fluorescence within the filtered wavelength of CY3 (Figure S3). The observed texture could have resulted from the tearing of GMA and embedded alginate during cross-sectioning. Such CY3 signals (i.e., false detection of anammox colonies) was successfully eliminated during data processing with size and circularity filtering because the autofluorescent textures were large non-circular objects along the direction of cutting.

The specificity of FISH probes was verified using a pure culture ofNitrosomonas and an anammox enrichment culture. FISH probes successfully detected AOB and anammox bacteria in both suspended and encapsulated samples (Figure S4). No NOB colonies were detected in the granular, non-encapsulated anammox enrichment culture. This is probably because low concentrations of NOB were present in the culture, which could have resulted in no observable colonies in the 5 μm² area under the microscope. Successful detection of NOB colonies was demonstrated in experiments with encapsulated biomass by both qPCR and FISH (Figure S5), indicating that NOB must have been present in the initial culture, just at a low density (see below).

The reproducibility of the cross-section results was examined by comparing three consecutive 10-µm sections from the same alginate bead. The sample was collected after 5 days of aerobic incubation of encapsulated Nitrosomonas . Cross-sections 100 µm apart were collected for the entire bead. Both the area and count of FISH-detected AOB colonies increased with depth from the surface from 0 to 300 µm, followed by a decrease as one moved further toward the center of the bead (1500 µm), indicating preferred growth of AOB near the surface (Figure 2 ). One would have expected the microbial counts and areas to be roughly symmetrical around the core of the bead (e.g., 0 to 1500 µm mirrored in the 1500 to 3000 µm slices). This, however, was not the case. Indeed, the quality of the sections obtained worsened as the slices were taken from deeper than the core of the bead. The integrity of the slices at large depths (>1500 µm) could therefore not be used for analysis. This was thought to be caused by the fact that the GMA embedding material was molded into the conical shape of the microcentrifuge tube (Figure S6). At the tip of the tube, the slices consisted primarily of the bead itself, but as the cross-sectioning went deeper into each sample, the slices consisted of more of the GMA surrounding the bead. It was thought that as a result, greater friction occurred between the GMA/sample and the blade, compared to the sample only and the blade, resulting in tearing and poor sample quality. To ensure the quality of our analysis, we thereafter performed cross-sectioning only up to the center of each alginate bead sample. This is an important limitation of this technique, as it is predicated on the assumption that microbial growth is symmetrical about the core of a bead. If investigating other embedded geometries or systems where growth might not be symmetrical (larger beads in a high-rate packed bed for example), more samples of embedded microorganisms will need to be analyzed to obtain a clear picture of growth.

When one looks at the 0-1500 µm depths, the largest and densest AOB colonies were observed around 300 µm from the surface of the alginate bead. The areas of AOB colonies were statistically compared among the three neighboring sections, with the three consecutive sections at depthD ± 10 µm treated as technical triplicates representing a single alginate bead at depth D . No significant differences were observed for 11 sets of sections (Kruskal-Wallis test, p > 0.05), demonstrating that these three neighboring slices could indeed act as replicates. Nevertheless, four sets of slices, out of 15, showed significant differences among the three “replicate” slices (at 300, 400, 1100, and 1200 µm). Part of this appears to be the result of the larger colony areas and larger variations in colony area observed at depths 300 and 400 µm where most of the growth occurred.